US 8139878

I'll research this patent through web searches focusing on the USPTO record and CAFC 2026 dockets.

Let me search more specifically for Federal Circuit appeals involving this patent and the ITC determination.

Let me do one final check on any Federal Circuit appeal from the ITC investigation and confirm the USPTO bibliographic record.

I have enough to provide the requested summary. Here are my findings, with explicit notes on uncertainty where applicable.

CAFC 2026 docket search result

No Federal Circuit docket for 2026 specifically tied to U.S. Patent 8,139,878 surfaced in my searches. I checked USPTO/patent records and CAFC-related coverage for 2026 (and surrounding years). The CAFC appeals involving VideoLabs, Inc. that do appear in 2024–2026 dockets (e.g., VideoLabs, Inc. v. Netflix, Inc., No. 25-1131; VL Collective IP, LLC v. Netflix, Inc.; Haynes Boone's Feb. 2026 affirmance in a Unified Patents PTAB appeal) concern other patents (7,440,559; 8,605,794, etc.) — not 8,139,878. I did not find a CAFC appeal of the ITC's 2023 invalidity determination on this patent. If a 2026 CAFC docket exists, it was not locatable through these searches, so I flag that as an unresolved item rather than fabricate one.

US Patent 8,139,878 B2 — Bibliographic summary

Notable litigation/validity event (non-CAFC): In ITC Investigation No. 337-TA-1323 (Certain Video Processing Devices and Products Containing the Same), VideoLabs asserted claims 1–4 of the '878 patent against Acer, ASUS, Lenovo, MSI, and Motorola. On August 1, 2023, the Commission affirmed (with modifications) the ALJ's summary determination that the asserted claims are invalid for obviousness-type double patenting and terminated the investigation with a finding of no Section 337 violation (88 FR 52209). The patent also appears in numerous district-court cases (W.D. Tex., D. Del., E.D. Tex.) that were mostly settled or dismissed (e.g., VideoLabs v. Dell, 1:23-cv-01366, dismissed with prejudice Jan. 2024; VideoLabs v. TCL, 2:25-cv-00161, E.D. Tex.). I found no indication these were appealed to the CAFC in 2026.

Abstract (plain-language paraphrase)

The patent is directed to picture coding and picture decoding performed block-by-block after transforming an image into coefficients representing spatial frequency components. To improve coding efficiency regardless of picture quality, the encoder predicts the number of non-zero coefficients in the current block from the non-zero coefficient counts of already-coded neighboring blocks, selects variable-length-coding (VLC) tables based on that predictive value, and then variable-length-codes the current block's non-zero coefficient count using the selected table(s). A corresponding decoder performs the mirror-image prediction, table selection, and variable-length decoding using counts from already-decoded neighboring blocks. In the prior art (e.g., MPEG-4, ISO/IEC 14496-2), only a single VLC table was used per prediction mode, causing efficiency to vary with picture content; this invention adapts the tables to local content.

Independent claims — plain-language overview

Caveat on claim text: The excerpt of the patent supplied to me contains the specification/description but not the verbatim claims section, so I cannot quote the independent claims word-for-word with complete confidence. Based on the disclosed claim language mirrored in the specification, the independent claims are method claims for (1) coding and (2) decoding (with device/program claims contemplated in the disclosure as further independent claims). The ITC record confirms claims 1–4 were the asserted claims.

Independent claim 1 — Picture coding method (representative):

1. A method of coding an image on a block-by-block basis by transforming the image into spatial-frequency coefficients.
2. Predicting: calculating a predictive value of the number of non-zero coefficients in the current block based on the non-zero coefficient counts of coded blocks located on the periphery of the current block (e.g., blocks above and to the left).
3. Table selecting: selecting variable-length-coding table(s) based on that predictive value.
4. Variable-length coding: coding the number of non-zero coefficients in the current block with reference to the selected table(s).

Independent decoding-method claim (mirror image): Same steps on the decoder side — predicting the current block's non-zero coefficient count from already-decoded peripheral blocks, selecting variable-length-decoding table(s) based on the prediction, and decoding the received variable-length code into the non-zero coefficient count using the selected table(s).

Notable dependent-claim concepts (from the claim-language definitions in the specification): predictive value computed as an average (or max/min/median) of neighbor counts; selection among multiple VLC tables and/or code tables (number-of-coefficients → code number → VLC code); boundary handling — predictive value 0 at picture/slice boundaries; using only the left neighbor's count, only the above neighbor's count, or their average depending on which boundaries the current block abuts; storing neighbor counts until the current block is coded/decoded; and encoding/decoding a difference value between the predictive value and the actual count as an alternative.

Key uncertainties

• Exact verbatim claim 1–4 text — not in the excerpt; reconstructed from the specification's mirrored claim language and the ITC's reference to "claims 1–4."
• One-day discrepancies in filing date (Oct. 24 vs. 25, 2007), priority date (Apr. 14 vs. 15, 2002), and grant date (Mar. 19 vs. 20, 2012) between Google Patents and Unified Patents data; I used the Google Patents record (authoritative full text supplied) as primary.
• No CAFC 2026 docket for this patent was found; the most recent substantive validity event remains the ITC's August 2023 obviousness-type-double-patenting determination, which I could not confirm was appealed.

I'll research litigation involving US Patent 8,139,878 across multiple sources, prioritizing live search results over training data.

I have strong initial results. Let me now verify specifics on the Unified Patents caselist, the ITC investigations, and the most recent 2025–2026 cases to confirm which assert patent 8139878 and their current status.

I have reached the search-step limit, but I have gathered sufficient verified data from the supplied patent-page litigation record, USPTO/PTAB-related filings, court documents, and live search results to compile the litigation list. I'll flag confidence levels for each item.

Litigation involving U.S. Patent 8,139,878 ("'878 patent")

Nearly all litigation has been brought by VideoLabs, Inc. (with co-plaintiff VL Collective IP LLC in some cases), the current assignee, asserting the '878 patent alongside related codec patents (commonly 7,769,238; 8,208,542; 7,372,452; 7,970,059, etc.). There is no litigation naming the '878 patent as defendant-side (e.g., a declaratory-judgment action against the patent owner) that I can confirm; the cases below are infringement actions and one ITC investigation.

A. District-court infringement actions (VideoLabs as plaintiff)

2021–2022 wave (W.D. Tex. — Judge Albright)

2023–2026 wave (W.D. Tex. and E.D. Tex.)

B. ITC investigations

C. PTAB proceedings

Key caveats

• Highest-confidence items (verified against primary/contemporaneous documents): W.D. Tex. 6:22-cv-00719 (Acer), 6:22-cv-00721 (Lenovo/Motorola), 6:22-cv-01086 (HP), D. Del. 1:22-cv-00229 (Netflix), 1:22-cv-00680 (Meta), W.D. Tex. 1:23-cv-01366 (Dell), E.D. Tex. 2:25-cv-00161 (TCL), and ITC 337-TA-1323.
• Not fully verified: the current (2026) status of the Amazon (6:22-cv-00079), ASUS (6:22-cv-00720), MSI (6:22-cv-00722), Gigabyte (2:25-cv-00704), and the 2026 E.D. Tex. case (2:26-cv-00232); the Dell 6:21-cv-00456 case details; and ITC matters 337-TA-1341 and 337-TA-3650 (the latter's number is suspect).
• Outcome summary: The '878 patent survived early § 101 motions (Netflix, Meta) but suffered its most significant adverse event in ITC 337-TA-1323 (Aug. 1, 2023) — asserted claims 1–4 held invalid for obviousness-type double patenting, with no Section 337 violation found. District-court cases have largely resolved by settlement/stipulated dismissal (Acer, HP, Dell) or remain in earlier stages (TCL, Gigabyte, 2026 case). I found no Federal Circuit appeal of the ITC invalidity determination as of the current date.

I'll research the PTAB proceedings on this patent. The structured data says the USPTO ODP API has no AIA trials on file, but the Google Patents record itself references IPR2022-00629 — so let me verify what actually exists.

Good initial results. I've identified two IPRs so far (IPR2022-00629 by Dell and IPR2018-00658 by Huawei), both institution-denied. Let me dig deeper into both and check for any other proceedings.

I now have solid data on both IPRs. Let me verify whether IPR2018-00658 targeted the '878 or the related '238 patent (sources conflict), check the rehearing outcome in IPR2022-00629, and confirm no other PTAB proceedings or reexaminations exist.

I've hit my search limit, but I have sufficient verified material to answer. One important contradiction to flag up front: the structured ODP block in this prompt says "no AIA trial proceedings on file," but that is stale/incomplete — the Google Patents record for US 8,139,878 itself links to PTAB case IPR2022-00629, and that proceeding is well documented across PTAB paper copies (Banner Witcoff), Patexia dockets, and press coverage. Details below.

Proceedings overview

For US 8,139,878 B2 there is one (1) verified AIA trial proceeding — IPR2022-00629 ([Dell Technologies Inc. et al](/litigations/by-defendant/Dell%20Technologies%20Inc.%20et%20al). v. VideoLabs, Inc.) — and its status is institution denied (2022-10-04); there are zero PTAB proceedings with claims invalidated, zero sustained by Final Written Decision, zero settled, and zero active. A second proceeding, IPR2018-00658 (Huawei v. Optis), is frequently cited in this patent's context but the USPTO's own docket identifies its target as the sibling patent 7,769,238, not the '878 — see the flag in that section. Bottom line for a defendant: the PTAB has never canceled a claim of the '878, because no petition ever got past the institution stage — but the claims were nonetheless held invalid by the ITC (not the PTAB) in 2023 for obviousness-type double patenting. The PTAB record alone gives you no canceled claims to lean on; the ITC determination is your strongest validity weapon, and it is not an IPR-derived win.

IPR2022-00629 — Dell Technologies Inc. & Dell Inc. v. VideoLabs, Inc.

• Type: Inter Partes Review
• Filed: 2022-03-07 (petition papers dated 2022-03-07; related Dell petitions IPR2022-00628 on the '238 and IPR2022-00701 on the '059 were filed 2022-03-02 and 2022-03-17 respectively and decided concurrently)
• Status: Institution Denied (verbatim from Patexia docket); plain-English gloss: the Board never instituted trial, so there is no Final Written Decision and no claim was canceled
• Judge panel: Jeffrey S. Smith (author of the decision), joined by Karl D. Easthom and Patrick M. Boucher
• Petition grounds: Claims 1, 2, 3, and 4 of the '878 challenged under § 103 obviousness, relying on references including ITU-T H.324 (Ex. 1004), ITU-T H.263 (Ex. 1005), Bjontegaard (US 7,099,387, Ex. 1006), JVT-C167 (Ex. 1007), and Tanaka (US 6,741,656, Ex. 1010), with a declaration from Gary Sullivan (Ex. 1011). The decisive flaw: Dell's primary references were dated May and August 2002 — after the priority date Dell had to beat. (I could not verify the exact ground-by-ground claim/reference matrix within my search limits; the above is from the exhibit list and contemporaneous press summaries.)
• Institution decision: Denied — 2022-10-04, Paper 10. The Board held that the Japanese priority application JP 2002-112665 (filed 2002-04-15) provides adequate written description support for the challenged claims (including the audio-processing/demultiplexing limitations Dell argued lacked support), entitling the '878 to the 2002-04-15 priority date. Because Dell's primary references post-dated that date, they are not prior art. The Board's language, as reported: "This priority showing is fatal to the Petition's obviousness showing." See the decision copy at bannerwitcoff.com (https://bannerwitcoff.com/wp-content/uploads/2022/10/IPR2022-00629.pdf).
• Final Written Decision: None — trial was never instituted.
• Settlement / termination: No settlement; the case terminated administratively by virtue of the institution denial. Petitioner filed a Request for Rehearing (Paper 11, 2022-10-26) and a Precedential Opinion Panel request (Paper 12, 2022-10-27); the outcome of those requests is not verified in my searches and should be confirmed on PTAB E2E (Papers 13+). Do not assume they were granted — Board rehearing of institution denials is rarely granted.
• Appeal: None found. Institution denials are not appealable under 35 U.S.C. § 314(d), and no CAFC docket for this proceeding surfaced.
• Defensive value: Limited but real. You cannot cite this as a merits win — the Board never reached the prior art. What it does tell you: the gating fight on this patent is priority date (2002-04-15) and written-description support in the Japanese application for the audio-processing limitations. Any new IPR must use art that predates 2002-04-15, or must first win the written-description battle Dell lost.

IPR2018-00658 — Huawei Device Co., Ltd. v. Optis Wireless Technology, LLC (related-proceeding flag — target is the sibling '238 patent, not the '878)

• Type: Inter Partes Review
• Filed: 2018-02-16
• Status: Institution Denied (2018-10-09, Paper 14); Petitioner's rehearing request denied 2019-04-03 (Paper 18); Patent Owner's rehearing request denied as moot (Paper 19)
• Judge panel: Miriam L. Quinn, Patrick M. Boucher, and a third APJ (name truncated in the source I retrieved)
• Petition grounds: Challenge to claim 1 (per the decision documents)
• Institution decision: Denied 2018-10-09
• ⚠️ Accuracy flag — do not cite this as a '878 proceeding without checking: The PTAB's own later decision in IPR2022-00629 (quoting the parties' mandatory notices) describes IPR2018-00658 as "denying institution on petitioner's challenge to claim 1 of the '878 patent." However, the USPTO's own documents in IPR2018-00658 — including the petition, the institution decision, and rehearing orders — identify the patent as US 7,769,238 B2 (application 11/976,758), and Patexia/IPverse/National Law Review all list 7,769,238 as the respondent patent. The '878 is application 11/976,551. The preponderance of primary evidence says IPR2018-00658 targeted the '238 (a same-specification sibling), and the "'878" characterization in the IPR2022-00629 decision appears to be an error inherited from a party filing. If you intend to rely on IPR2018-00658 as a '878 proceeding, pull Paper 14 from PTAB E2E first.
• Defensive value: If it is (as the primary record shows) a '238-patent proceeding, it is irrelevant to the '878 except as family context showing that every AIA challenge to this patent family has been stopped at institution — Huawei in 2018, Dell in 2022.

Strategic summary

Claims CANCELED vs. SUSTAINED vs. UNTESTED — the PTAB record: No claim of the '878 has ever been canceled by the PTAB, and no claim has ever been sustained in a Final Written Decision either, because no IPR on the '878 has ever been instituted. All claims that appear in the litigation/IPR record — claims 1–4 (the claims Dell challenged and the claims VideoLabs asserted at the ITC) — are therefore UNTESTED on the merits at the PTAB. Every claim remains legally in force as far as the PTAB is concerned, and the only administrative merits finding against them came from a different forum: the ITC in Inv. No. 337-TA-1323, which on 2023-08-01 affirmed (with modifications) a summary determination that asserted claims 1–4 of the '878 are invalid for obviousness-type double patenting, terminating the investigation with no Section 337 violation (88 FR 52209, published 2023-08-07). That ODP holding is not an IPR ground and did not cancel the claims at the USPTO — the patent's Google Patents legal status still reads "Active."

Estoppel landscape: § 315(e)(2) estoppel attaches only after a Final Written Decision on the merits. Because every petition against this patent — IPR2022-00629 on the '878, and IPR2018-00658/IPR2022-00628 on the '238, and IPR2022-00701 on the '059 — was denied at institution, no petitioner is estopped. Dell, its privies, and any new challenger remain free to raise every § 102/§ 103 ground they raised or could have raised. The practical constraint is not estoppel but priority: the Board has twice (2018 and 2022) credited the 2002-04-15 Japanese priority date for this family, so a viable IPR needs (a) prior art dated before 2002-04-15, or (b) a successful attack on written-description support in JP 2002-112665 for the claimed audio-processing limitations — the exact argument Dell lost.

Pattern signals: Dell ran a coordinated three-petition campaign (IPR2022-00628/00629/00701) across the '878 and its two same-family siblings, and all three were denied the same day on the same priority rationale — a sign this family has proven institution-proof so far, not that its claims are strong on the merits. Unified Patents is present in the ecosystem as a data portal tracking the case and as an established antagonist of VideoLabs on other patents (winning IPR invalidation of VideoLabs's '794 patent in December 2023 and a reexam grant on the '790), but Unified itself has not filed an IPR against the '878 as far as I can verify. VideoLabs/Optis has successfully defended the family at the PTAB institution stage every time; its losses have come in the ITC (the 2023 ODP ruling) and in district-court/PTAB fights over other VideoLabs patents (e.g., Netflix's invalidations of the '790, '559, and '794).

Recommended next steps

1. Lead with the ITC determination, not the PTAB record. If you are a defendant facing claims 1–4 of the '878, the strongest existing invalidity holding is the Commission's 2023-08-01 determination in Inv. No. 337-TA-1323 affirming that claims 1–4 are invalid for obviousness-type double patenting (Federal Register notice 88 FR 52209: https://www.federalregister.gov/documents/full_text/xml/2023/08/07/2023-16773.xml). An ITC determination is not binding on a district court, but it is a fully developed agency merits decision you should put in front of the court on day one, and ODP is a defense you can raise directly in district court (it is not available in an IPR, so the PTAB is the wrong forum for that ground).
2. Check the term/expiration threshold before anything else. The adjusted expiration shown on the Google Patents record is 2026-06-04, and today's date (2026-09-03) is after that date. Verify the exact expiration date and any terminal disclaimers — if the patent has expired, recovery is limited to pre-expiration conduct, which sharply caps exposure and may moot injunctive relief entirely.
3. If you still want an IPR: confine yourself to art published or patented before 2002-04-15, and budget for the written-description fight over the Japanese priority document that killed Dell's petition (IPR2022-00629, Paper 10). Confirm the status of Dell's rehearing/POP requests (Papers 11–13+) on PTAB E2E before assuming the priority issue is fully settled — but note that even a rehearing grant would only have reached institution, not a merits FWD.
4. Confirm the IPR2018-00658 target patent (7,769,238 vs. 8,139,878) directly from Paper 14 on PTAB E2E if you plan to cite it; the primary USPTO docket says '238, and citing it as a '878 proceeding could draw a sanctions-level accuracy challenge from opposing counsel.

Key sources: IPR2022-00629 institution decision (Paper 10) at https://bannerwitcoff.com/wp-content/uploads/2022/10/IPR2022-00629.pdf; Patexia docket at https://services.patexia.com/lawsuits/Dell-Technologies-Inc-et-al-v-VideoLabs-Inc-id-[162370](/patent/162370); Unified Patents case page at https://portal.unifiedpatents.com/ptab/case/IPR2022-00629; IPR2018-00658 docket at https://services.patexia.com/lawsuits/Huawei-Device-Co-Ltd-et-al-v-Optis-Wireless-Technology-LLC-et-al-id-[111708](/patent/111708)/documents; ITC termination notice at the Federal Register link above.

I'll research the assignment record for US 8,139,878 through available sources before reconstructing the chain.

Good — I found some reel/frame data. Let me drill into the specific recordings for this patent.

Let me try to pin down the specific reel/frame numbers for this patent's key recordings.

Assignment / Ownership Chain — U.S. Patent 8,139,878 B2

Coverage note (read first): The USPTO's Assignment Center per-patent view did not return structured reel/frame detail in the sources I could reach; what I could reach is (a) the patent's own legal-events assignment records (which Google Patents ingests from the USPTO assignment database, so they are recordings, not corporate announcements), and (b) a Rule 1.324 petition filed in a sibling Panasonic–Optis patent that quotes an actual reel/frame. I therefore report reel/frame only where I have a source, and mark the rest as not retrieved rather than guess. Correspondent-of-record data — the field the task asks me to prioritize — was not retrievable for any of these recordings; flagged in each entry and in Signal 3.

Inventors

All four inventors are the named assignors on the assignment of inventors' interest to Matsushita Electric Industrial Co., Ltd. (recorded 2017-11-12 per the patent's legal-events record), and Matsushita Electric Industrial is the entity that renamed itself Panasonic Corporation on 2008-10-01. That assignment is direct evidence of a single-employer inventive team, not a joint venture or university/industry collaboration.

Unusual patterns: None of the classic "all inventors gone within 12 months" fire-sale tell is present. There is no evidence in the record of any inventor departing the original assignee, and no reissue or inventorship-correction event on the '878 itself. However, there is an unusual chain-of-title timing anomaly: the inventors' assignment to their employer was recorded in November 2017 — roughly 5 years after the 2012-03-20 issuance and 3.7 years after Panasonic had already sold the patent to Optis (2014-02-23). That is a belated perfection of an originally unrecorded (or mis-recorded) employment assignment, most plausibly executed in the 2002 era but recorded only during a portfolio title-cleanup. It is a defect-cure, not an arm's-length transfer.

Original assignee

Panasonic Corporation (the entity named on the face of the issued patent; formerly Matsushita Electric Industrial Co., Ltd., renamed 2008-10-01). The application itself was filed 2007-10-25 by Panasonic Corp.

• Primary line of business: consumer and industrial electronics — video codec ICs/LSIs, camcorders, TVs, AV equipment. This family sits squarely in Panasonic's H.264/video-coding development, and the sibling '238 patent was later asserted against H.264 CAVLC implementations.
• Did they ship a product embodying the claims? Very likely in the sense that matters: the family's own litigation record shows VideoLabs asserting the sibling '238 and the '878 against H.264 CAVLC entropy decoding in third-party TVs/set-top boxes (Hisense notice letter, 2022-12-27: "'238 … owned by VideoLabs, Inc., and specifically claim no. 1 with respect to the H.264 CAVLC entropy decoding employed by Hisense's products"). Panasonic was itself an H.264/JVT participant, so the claimed technique was implemented in Panasonic's own codec products/standards participation. I have not independently verified a specific Panasonic product shipping an accused configuration.
• Current status: Operating. Panasonic Corporation still exists under a holding-company structure as Panasonic Holdings Corporation (effective 2022-04; Panasonic Holdings Corp is in fact listed as a "Parent Company" on Unified Patents' page for this patent). No bankruptcy — the portfolio exit was a voluntary divestiture, not a distress sale.

Assignment timeline

All dates below are the recording dates shown in the patent's assignment record (execution dates may precede them and are not always exposed). "Not retrieved" = Assignment Center did not yield the field in the sources I could reach; no reel/frame has been invented.

• Recorded 2017-11-12 (execution date not shown; document is 2002-era) — Reel 044748/0001 (candidate match only — see note)

  • Conveyance: Assignment of inventors' interest
  • Assignor: Abe, Kiyofumi; Hagai, Makoto; Kadono, Shinya; Kondo, Satoshi
  • Assignee: Matsushita Electric Industrial Co., Ltd. (pre-2008 name of Panasonic Corporation)
  • Correspondent: not retrieved
  • Context: Belated perfection of the original employment assignment — a chain-of-title cleanup recorded 13 years after execution and 3.7 years after the same patent had already been conveyed to Optis.
  • Reel note: an INPADOC legal-status entry for a different Panasonic-origin patent shows a Panasonic-side recording at 044748/0001 dated 2017-11-13. I could not confirm this is the '878's reel. Treat as a lead, not a fact.

• Recorded 2014-01-24 — Reel not retrieved

  • Conveyance: Lien (grant of security interest)
  • Assignor: Optis Wireless Technology, LLC
  • Assignee: Highbridge Principal Strategies, LLC, as collateral agent
  • Correspondent: not retrieved
  • Context: Securitization — Optis pledged its patent portfolio as collateral for acquisition financing. Note the sequencing: the lien is recorded one month before Optis actually took title to the '878, i.e. an after-acquired-property pledge put in place ahead of the Panasonic closing.

• Recorded 2014-02-23 — Reel 032326/0707 (verified for a sibling patent recorded the same day; per-patent confirmation not obtained for the '878 — see note)

  • Conveyance: Assignment of assignors' interest
  • Assignor: Panasonic Corporation
  • Assignee: Optis Wireless Technology, LLC
  • Correspondent: not retrieved
  • Context: Portfolio divestiture / transfer-to-asserter — a bulk, multi-patent recordation. Corroborated independently: a Rule 1.324 petition in sibling patent US 8,064,919 states "an Assignment from PANASONIC CORPORATION to OPTIS WIRELESS TECHNOLOGY, LLC, recorded in the U.S. Patent and Trademark Office on February 23, 2014 at Reel/Frame 032326/0707" — the same recording date as the '878's Panasonic→Optis event. Optis' address of record in that petition is P.O. Box 250649, Plano, Texas 75025.
  • Reel note: I verified 032326/0707 exists and is dated 2014-02-23, but for the '919, not the '878. The '878 sits in the same same-day bulk recordation and may be on the same or a neighboring reel. Do not cite 032326/0707 as the '878's reel without pulling the per-patent Assignment Center record.

• Recorded 2014-03-12 — Reel not retrieved

  • Conveyance: Security interest
  • Assignor: Optis Wireless Technology, LLC
  • Assignee: Wilmington Trust, National Association (as collateral agent/trustee)
  • Correspondent: not retrieved
  • Context: Securitization — a second, layered encumbrance recorded 17 days after the Panasonic closing; consistent with a refinancing or a rated note issuance over the acquired portfolio.

• Recorded 2016-07-15 — Reel not retrieved

  • Conveyance: Release by secured party
  • Assignor: HPS Investment Partners, LLC (successor to Highbridge Principal Strategies)
  • Assignee: Optis Wireless Technology, LLC
  • Correspondent: not retrieved
  • Context: Securitization unwind — release of the 2014-01-24 Highbridge lien only.

• Recorded 2020-11-16 — Reel not retrieved

  • Conveyance: Release by secured party
  • Assignor: Wilmington Trust, National Association, as collateral agent
  • Assignee: Optis Wireless Technology, LLC
  • Correspondent: not retrieved
  • Context: Securitization unwind preceding sale — the 2014-03-12 Wilmington Trust encumbrance is cleared roughly five weeks before Optis sells to VideoLabs. Note the assignee field here reads "Optis Wireless Technology" without the "LLC" suffix — a naming slip in the record; there is no evidence of a distinct "Optis Wireless Technology" entity.

• Recorded 2020-12-22 — Reel not retrieved

  • Conveyance: Assignment of assignors' interest
  • Assignor: Optis Wireless Technology, LLC
  • Assignee: VideoLabs, Inc.
  • Correspondent: not retrieved
  • Context: Transfer-to-asserter — the operative NPE hand-off, and the last link in the chain. VideoLabs asserts in its D. Del. pleadings that "VL is the assignee and lawful owner of all right, title, and interest in and to the '878 Patent."

Net chain: Inventors → Matsushita (recorded late) → Optis Wireless Technology, LLC → VideoLabs, Inc. Two LLC hops, one operating company, three encumbrance/release cycles.

Timeline diagram

timeline
    title Ownership of US 8139878
    2002 : Inventors assign to Matsushita
    2007 : Application filed by Panasonic
    2012 : Patent issued
    2014 : Highbridge lien recorded
         : Panasonic assigns to Optis Wireless
         : Wilmington Trust security interest
    2016 : HPS releases lien
    2017 : Inventor assignments recorded late
    2020 : Wilmington Trust releases encumbrance
         : Optis assigns to VideoLabs Inc
    2021 : First VideoLabs suit on this patent

NPE / troll-pattern signals

1. Shell-entity transfer — PRESENT (moderate).
The patent moved from an operating manufacturer to a licensing-only entity in 2014 (Panasonic → Optis Wireless Technology, LLC, recorded 2014-02-23). Concrete corroboration beyond the name: (a) Optis' address of record is a P.O. Box — P.O. Box 250649, Plano, Texas 75025 (quoted in the sibling Rule 1.324 petition); (b) Optis' patent assets were pledged to a hedge fund/credit manager as collateral (Highbridge lien 2014-01-24; Wilmington Trust security interest 2014-03-12) — the portfolio was the asset, which is not how an operating company's patents are typically held; and (c) Optis appears in the record only as a patent plaintiff/licensor. Caveat on the second hop: VideoLabs, Inc. is not a two-person shell — its own pleading describes it as "founded in 2018 as part of an industry-sponsored and -funded effort," and its 2022 demand letter advertises "a significant patent portfolio (of over 125 hand-selected patents)." So this signal fires on the Optis link, not on VideoLabs.

2. Known asserter in the chain — PRESENT (strong).
Both intermediate and current owners are repeat, high-frequency patent plaintiffs.

• Optis Wireless Technology, LLC / PanOptis Patent Management, LLC: asserted Panasonic- and Ericsson-origin portfolios in Optis Wireless v. Huawei (2:17-cv-00123, E.D. Tex.), PanOptis v. Apple (E.D. Tex., $506.2M damages sought/obtained per Apple's proposed findings), Optis v. ZTE (2:15-cv-300), and Optis v. Kyocera/BlackBerry/ZTE (2:16-cv-00059).
• VideoLabs, Inc.: asserted this patent (and siblings) against Dell, Amazon, Netflix, Meta, Acer, ASUSTeK, Lenovo/Motorola, MSI, HP, TCL, GigaByte, and Hisense, plus ITC Inv. Nos. 337-TA-1323 and 337-TA-1341. A Taiwan TIPO (Intellectual Property Office) report expressly classifies VideoLabs as a type of NPE ("廣義上屬NPEs之一種類型"). Unified Patents tracks the patent and IPR2022-00629 on its portal.
• Neither entity appears on the enumerated classic rosters (Acacia, Marathon, IV, Wi-LAN, etc.), so I flag this as present by conduct (high-frequency plaintiff + NPE characterization), not by roster membership.

3. Repeat correspondent across the chain — UNCLEAR / NOT DETERMINABLE.
This is the signal I was asked to prioritize and it is the one I could not verify. I retrieved no correspondent-of-record name for any of the eight recordings above, and I will not supply one by inference. What I can report is structural: the 2014-02-23 Panasonic→Optis conveyance was a single-day bulk recordation across a multi-hundred-patent portfolio (evidenced by the sibling '919 sharing the same recording date, reel 032326/0707), which is the signature of one law firm/docketing vendor running a portfolio-level filing rather than per-patent counsel. Action item: the correspondent field is visible on the Assignment Center per-patent view and on the cover sheet of each reel/frame copy — pull reels 032326/xxxx (2014) and the Dec-2020 recording and check whether the same firm/attorney recurs across the Optis lien, the Wilmington Trust security interest, and the VideoLabs assignment. That recurrence test is the whole point of the signal and it is unanswered on the current record.

4. Cascading transfers through chained LLCs in <24 months — NOT PRESENT.
There is exactly one LLC-to-LLC hop (Optis, 2014 → VideoLabs, 2020), separated by 6.8 years. What does cluster is encumbrances: three recordings within seven weeks in early 2014 (lien 2014-01-24; assignment 2014-02-23; security interest 2014-03-12). That is an acquisition-financing cluster, not an LLC cascade, and the assignees are a credit manager and a trustee rather than affiliated shells. Marked not present as defined.

5. Pre-litigation transfer — PRESENT (strong).
Assignment executed/recorded 2020-12-22 (Optis → VideoLabs); the first infringement action naming this patent is VideoLabs, Inc. v. Dell Technologies Inc., W.D. Tex. 6:21-cv-00456, filed 2021-05-03 (per the Patexia related-cases docket for the VideoLabs/TCL family). That is ~4.4 months from title transfer to first suit — inside the 6-month window. Consistent with a chain arranged to place clean standing in an asserting vehicle immediately before a filing campaign. (The '878 was not among the patents construed in the earlier Optis v. Huawei case, so its assertion history genuinely begins with VideoLabs.)

6. Bankruptcy fire-sale — NOT PRESENT.
Panasonic/Matsushita never filed Chapter 7/11 in connection with this portfolio. The 2014 exit was a solvent, negotiated divestiture of "charted/chartable" standard-relevant patents. (Portfolio-acquisition context from Apple's findings of fact: the Optis Wireless portfolio was "formed around a portfolio of patents originating from Ericsson and Panasonic," with portfolios "comprised of patents hand-picked from each predecessor-in-interest.")

7. Privateering — UNCLEAR (leans not present).
The first hop is a genuine operating-company-to-NPE transfer (Panasonic → Optis), and the second continues down the NPE chain (Optis → VideoLabs). But classic privateering requires the original operating company to retain an economic interest and/or direct the assertions against its competitors. I found no evidence of Panasonic retaining a royalty interest in, or directing VideoLabs' assertions of, the '878. On the current record this reads as a straight divestiture that fed an NPE chain, not privateering.

8. Defensive aggregator — NOT PRESENT.
The chain terminates at VideoLabs, Inc., an active plaintiff with a live 2026 docket (E.D. Tex. 2:26-cv-00232) — the opposite of a defensive terminus. Nothing in the chain touches RPX, AST, LOT Network, Unified Patents, or OIN.

Verdict

NPE — high confidence.

Three strong signals fire on specific recorded events: (i) known/high-frequency asserter in the chain — Optis Wireless Technology, LLC took title by assignment recorded 2014-02-23 and VideoLabs, Inc. by assignment recorded 2020-12-22, both serial patent plaintiffs on these exact codec patents; (ii) pre-litigation transfer — title moved to VideoLabs on 2020-12-22 and the first suit naming this patent was filed 2021-05-03 (Dell, 6:21-cv-00456), ~4.4 months later; and (iii) shell/licensing-only transfer with asset-backed collateralization — Panasonic's 2014 assignment to Optis, whose address of record is P.O. Box 250649, Plano, Texas, and whose portfolio was pledged to Highbridge Principal Strategies (lien, 2014-01-24) and Wilmington Trust (security interest, 2014-03-12) with releases in 2016-07-15 and 2020-11-16 immediately preceding the VideoLabs sale. The one gap is Signal 3: no correspondent of record could be retrieved for any link, so the "same lawyer runs every shell" test is unanswered rather than negative, and I flag it as the next thing to pull.

Verification / next steps

• Assignment Center (current UI): https://assignmentcenter.uspto.gov/ — search patent number 8139878; the per-patent view yields reel/frame, executed date, and the correspondent field the task asks for.
• Legacy Assignment Search: https://assignment.uspto.gov/patent/index.html
• Priority pulls: (1) reel 032326/xxxx series, recorded 2014-02-23 (Panasonic → Optis), to confirm the '878's exact frame and the submitting attorney; (2) the recording behind the 2020-12-22 Optis → VideoLabs assignment, to check whether the same correspondent also handled the 2014-01-24 lien and 2014-03-12 security interest; (3) the reel behind the 2017-11-12 inventor → Matsushita correction, to date the original execution and confirm the chain-of-title gap.
• Corroborating document already in hand: Rule 1.324 petition in sibling US 8,064,919 quoting the 2014-02-23 / Reel 032326/0707 Panasonic→Optis recording and Optis' Plano, TX P.O. Box address.

Accuracy flags on this deliverable

• Reel/frame numbers are the weakest verified data here. Only 032326/0707 is sourced, and it is sourced to the sibling '919 patent, not the '878. 044748/0001 is a lead from an INPADOC entry for a different patent. All other reel/frame fields are marked not retrieved. Do not populate a chain-of-title chart from these as if per-patent verified.
• Execution dates ≠ recording dates. Every date above is a USPTO recordation date as surfaced in the patent's legal-events record. The 2017-11-12 inventor→Matsushita document almost certainly was executed in 2002; the 2020-12-22 and 2014-02-23 dates are recordation dates and the underlying signatures may predate them.
• The patent's prior sections note its legal status was "Active, expires 2026-06-04" — i.e., at today's date (2026-09-24) the term has run. That does not change the ownership chain, but it materially changes the value of the chain: the owner of record, VideoLabs, Inc., currently holds a patent whose adjusted expiration has passed, so any remaining posture is backward-looking recovery only.

I found the cited-reference list beginning to surface. Let me pull the full "References Cited" from the Justia record and the Google Patents citation listing to make sure I capture every reference (U.S., foreign, and non-patent literature).

USPTO Record Verification — U.S. Patent 8,139,878 B2

Verified bibliographic record (consistent across the Google Patents full text supplied, Justia's USPTO-derived record, and the Unified Patents portal):

Status of the claims for this analysis: the asserted claims in ITC Inv. No. 337-TA-1323 were claims 1–4 (the ITC affirmed an obviousness-type-double-patenting invalidity finding over U.S. 7,769,238 on Aug. 1, 2023). The verbatim claims section is not in the excerpt supplied to me, so I analyze against the reconstructed claim features mirrored in the specification: (i) predicting the number of non-zero transform coefficients (Nc) of a current block from Nc of previously coded/decoded peripheral blocks; (ii) selecting VLC table(s) — optionally both a "code table" and a "VLC table" — based on the predictive value; and (iii) VLC-coding/-decoding Nc itself. Dependent concepts include averaging (or max/min/median), above-and-left neighbor use, boundary rules (predictive value 0 at picture/slice edges), and storing neighbor counts.

A. References Cited on the Face of the '878 Patent

The following is the "References Cited" list as printed on the '878 front page (source: Justia's USPTO-derived record; the visible listing is complete for U.S. documents and appears truncated at JP 10-191324 for foreign documents). Completeness caveat: Unified Patents reports 46 patent-art items and 7 non-patent-literature items; I could verify only the items below from the visible record. Additional Japanese documents and NPL items may exist on the printed front page that were not visible in my search snippets.

A.1 U.S. Patent Documents (32, per Justia listing)

A.2 Foreign Patent Documents (partially visible; listing truncated)

A.3 Non-Patent Literature

Unified Patents counts 7 NPL items; only one is verifiable from the supplied specification: ISO/IEC 14496-2:1999(E), Information technology — Coding of audio-visual objects — Part 2: Visual, Dec. 1, 1999, p. 119 § 7.4.1 (Variable length decoding) — cited by the patent itself as the background art showing a single VLC table per prediction mode. The other six NPL items are not visible in my sources and I will not speculate on their identity.

B. Which Cited References Are the "Most Relevant" Prior Art, and Claim-by-Claim Anticipation Potential

B.1 Threshold legal point (applies to every reference below)

Because the '878 is a pre-AIA patent (application filed before Mar. 16, 2013), § 102 applies in its pre-AIA form. If the Apr. 15, 2002 JP priority date is credited (as the PTAB credited it for sibling '238 in IPR2022-00628):

• § 102(b) statutory bar: reference must have been patented or published before Apr. 15, 2001 → only items in rows 1–19 of A.1, the EP/JP documents, and the ISO/IEC 14496-2 NPL qualify on their dates.
• § 102(a)/(e): reference must have been known/used/patented/published before Apr. 15, 2002, or (for U.S. patents) filed in the U.S. before that date. Items 20–32 qualify only if their U.S. filing dates pre-date Apr. 15, 2002 (I could not verify file-wrapper dates for most).
• If priority fails (effective date = Oct. 25, 2007), then every U.S. patent issued before Oct. 25, 2006 (all 32), US 2002/0037108, the foreign documents, and ISO/IEC 14496-2 become § 102(b) bar art.

I flag this because the family's own IPR history (IPR2018-00658 and IPR2022-00628, both denied) shows that the dispositive art disputes turned on exactly this priority-date question, not on the merits of the references.

B.2 Subject-matter relevance — with explicit confidence limits

Honest limitation: I could verify the bibliographic data of the cited references from the USPTO-derived records, but I did not have access to the full text of each of the 32 U.S. patents during this analysis. A defensible "which claim does it anticipate" conclusion requires element-by-element comparison against the actual claim text. What follows is a reasoned, provisional assessment grounded in (i) the family's litigation record, (ii) each reference's date and inventor/assignee context, and (iii) the reconstructed claim elements. Where I lack confidence, I say so.

Tier 1 — Strongest § 102 candidates among the cited references (content plausibly covering Nc-from-neighbors + table selection)

• US 5,512,952 (Iwamura, Apr. 30, 1996) and US 5,510,841 (Iwamura, Apr. 23, 1996) — 1990s image-coding patents in the same technical lane as the claimed subject matter (block transform coding with adaptive variable-length coding of coefficient information). Being pre-2001, they are clean § 102(b) art against all claims. I could not retrieve their full texts to confirm whether they disclose prediction of the number of non-zero coefficients from peripheral blocks — that element, not generic VLC of coefficients, is the claim feature the PTAB found missing in JVT-B045 (see below). Provisional: plausible against dependent claims directed to generic "select a VLC table / code table and VLC table" limitations; unproven against the independent-claim peripheral-prediction element.
• US 6,148,109 and US 6,532,306 (both Boon et al., 2000 and 2003) — predictive image coding by a Philips inventor with multiple table-selection disclosures. 6,148,109 (issued Nov. 14, 2000) is a § 102(b) candidate on its date; 6,532,306 (issued Mar. 2003) is only § 102(e) art if its U.S. filing pre-dates Apr. 15, 2002 (likely, given Boon's 1990s work — unverified). Same element-by-element caveat as above.
• US 5,719,986 (Kato et al., Feb. 17, 1998) — Panasonic-era moving-picture coding art, pre-2001 § 102(b) candidate. Content unverified.
• US 2002/0037108 A1 (Matsumoto, published Mar. 28, 2002) — the only cited U.S. document published in the narrow window between Apr. 15, 2001 and Apr. 15, 2002, i.e., it is the only cited reference that can be § 102(a)/(e) art without depending on an earlier U.S. filing date. It is therefore a uniquely important candidate for anticipating the independent claims if it discloses neighbor-based Nc prediction and VLC-table selection. Its content was not retrievable in my searches — flagged as the single highest-value document to pull from the file wrapper.
• JP 10-191324 (Jul. 1998), EP 0 876 058 (Nov. 1998), JP 4-100390/4-223773/5-347710/6-232765/7-107492 (1992–1995) — Japanese/EP counterparts in the block-VLC space; all are § 102(b) candidates on their dates. Full-text review required; contents not verified here.

Tier 2 — Relevant background / cumulative art (multiple-VLC-table coding of transform coefficients, but no demonstrated peripheral-Nc prediction)

• US 5,500,678 and US 5,563,593 (Puri, 1996) — Puri's mid-1990s coding-systems patents; § 102(b) candidates. Content unverified; likely relevant to "select among multiple VLC tables" but not proven against the peripheral-prediction limitation.
• US 5,576,765 / 5,668,599 / 5,777,679 (Cheney et al., 1996–1998) — video-decoder / VLC-decoding patents; § 102(b) candidates; likely relevant mainly to the decoder-side dependent claims and to coefficient-value VLC (which the '878 admits was conventional).
• US 5,974,184 (Eifrig et al., 1999) — intra-frame coding (RealNetworks lineage); § 102(b) candidate; content unverified.
• US 6,009,229 (Kawamura, 1999), US 6,020,923 (Hosaka et al., 2000), US 6,141,385 (Yamaji, 2000), US 6,292,588 (Shen et al., 2001), US 6,341,144 (Haskell et al., 2002), US 6,449,395 (Kobayashi, 2002), US 6,473,459 (Sugano et al., 2002) — miscellaneous block-transform VLC art spanning the 1999–2002 period; date-based § 102(b)/(e) status as tabulated above; none individually verified for the Nc-prediction element.
• US 6,625,211 (Etoh et al., 2003), US 6,658,157 (Satoh et al., 2003), US 6,654,421 (Hanamura et al., 2003), US 6,907,142 (Kalevo et al., 2005), US 6,980,732 (Suzuki, 2005), US 7,010,168 (Suh et al., 2006), US 6,614,845 (Azadegan, 2003) — post-priority-date issuances; only viable as § 102(e) art if their U.S. filings pre-date Apr. 15, 2002 (largely unverified; 7,010,168's 2006 issue date makes a pre-2002 filing unlikely, so it is probably not § 102 art at all if priority holds). These are the weakest candidates on dates but include later Panasonic/Samsung/H.264-era CAVLC-type disclosures that could matter if the priority claim falls.
• ISO/IEC 14496-2:1999 (MPEG-4 Part 2) — cited by the '878 as the admitted background art that used only a single VLC table per intra/inter mode. It is the strongest evidence that everything except "peripheral-block Nc prediction for table selection" was conventional, and thus it is the key reference for reading the independent claims narrowly — but it teaches away from the claimed multi-table selection, so it is not an anticipatory reference by itself.

Tier 3 — The elephant in the room: art NOT on the face of the '878

The family's IPR record (against sibling US 7,769,238, whose claims are materially the same technique) identifies the closest art as:

• WO 03/063501 (Karczewicz), "Coding Transform Coefficients in Image/Video Encoders and/or Decoders" — priority US 10/054,610 filed Jan. 22, 2002 and provisional 60/369,500 filed Apr. 2, 2002; published Jul. 31, 2003. Karczewicz discloses exactly the '878's core mechanism: Nc of a current block is VLC-coded using one of multiple VLC tables selected by a function (e.g., average or median) of the Nc values of neighboring blocks (left, top-left, top, top-right) on the periphery. It is not a § 102 bar reference if the Apr. 15, 2002 priority holds (its WO publication post-dates Apr. 15, 2002), but its Jan. 22, 2002 / Apr. 2, 2002 filings pre-date the '878's Apr. 15, 2002 priority — which is precisely why the IPR2022-00628 priority ruling was outcome-determinative for the family.
• JVT-B045 and US 7,099,387 (Bjontegaard) — used in IPR2018-00658; the PTAB found JVT-B045's "last coded block" prediction was temporal, not the claimed spatial "peripheral blocks" relationship. That ruling defines the one element the Board considered non-conventional: spatial, peripheral-block-based Nc prediction.
• US 7,769,238 — the '878's own sibling; the ITC invalidated the '878's asserted claims 1–4 for obviousness-type double patenting over the '238 (not a § 102 issue, but the operative validity holding).

C. Bottom-Line Answers to Your Specific Questions

1. USPTO search for 8139878: Confirmed the USPTO-derived record (US 8,139,878 B2; app. 11/976,551; filed Oct. 2007; priority Apr. 15, 2002; issued Mar. 20, 2012; examiner Kanjibhai Patel). No "similar-number" patent was substituted.

2. Most relevant prior art cited on the face of the patent, ranked:

   • US 2002/0037108 A1 (Matsumoto, Mar. 28, 2002) — most valuable § 102(a)/(e) candidate by date; content unverified — pull this from the file wrapper first.
   • US 6,148,109 / 6,532,306 (Boon), US 5,512,952 / 5,510,841 (Iwamura), US 5,719,986 (Kato), EP 0 876 058, JP 10-191324 — the only cited documents old enough to be § 102(b) bar art and plausibly in the adaptive-VLC-of-transform-coefficients lane.
   • US 5,500,678 / 5,563,593 (Puri) and the Cheney et al. trio — solid § 102(b) art but likely limited to multi-table coefficient VLC, i.e., relevant to dependent limitations, not proven against the peripheral-prediction element.
   • ISO/IEC 14496-2:1999 — the admitted background; supports reading the independent claims narrowly.

3. Which claims each potentially anticipates: For the reasons stated, I cannot responsibly assign a claim-by-claim anticipation verdict to each of the 32 U.S. references without their full texts and the verbatim claims of the '878 (which were not in the excerpt supplied). What I can say with high confidence:

   • No single on-face reference has been shown (in this analysis or in the litigated record) to disclose all elements of the independent claims, particularly the spatial peripheral-block Nc prediction combined with table selection for coding Nc itself — the element the PTAB found missing in the closest art (JVT-B045) and present in Karczewicz.
   • The dependent-claim features (averaging/max/min/median of neighbor counts; "code table + VLC table"; above-and-left neighbor selection; boundary value 0; storing neighbor counts) are individually the most likely to be anticipated by the older cited art (e.g., Boon, Iwamura, Puri, Cheney, MPEG-4's own VLC apparatus), assuming each reference's full text supports the feature.
   • The independent claims 1–4 (per the ITC's asserted set) are most vulnerable to Karczewicz (WO 03/063501) — which is not on the face of the '878 — contingent on defeating the Apr. 15, 2002 priority date, and to US 2002/0037108 if its content proves out.

Recommended next step if you need a definitive § 102 chart: retrieve (a) the verbatim claims of the '878 from the USPTO full-text, (b) the file wrapper to confirm the effective filing date and the examiner's § 102 grounds, and (c) full texts of US 2002/0037108, US 6,148,109/6,532,306, US 5,512,952, and JP 10-191324. I can then produce the element-by-element anticipation matrix you're asking for.

§ 103 Obviousness Analysis — U.S. Patent 8,139,878 B2 ("Picture Coding Method and Picture Decoding Method")

Scope note on the "Prior Art section": The fetched patent text from Google Patents does not reproduce a discrete "Citations / Prior Art" section (the face-of-patent "References Cited" list was not extractable from the page dump). I therefore ground this analysis in the prior-art record that is actually documented for this patent: (1) the background art the specification itself admits (ISO/IEC 14496-2, MPEG-4 Visual, § 7.4.1, Dec. 1999); (2) the reference set Dell used in IPR2022-00629 against this very patent (ITU-T H.324, ITU-T H.263, Bjontegaard U.S. 7,099,387, JVT-C167 H.264 Joint Committee Draft, U.S. 6,741,656 (Tanaka)); and (3) the contemporaneous JVT/CAVLC standardization record (JVT-C028 RealNetworks; JVT-C088 Nokia/Matsushita) discussed in the VideoLabs v. Amazon claim-construction briefing. Where a document's date or disclosure cannot be verified, I say so explicitly.

1. Claims at issue and element breakdown

The asserted claims are claims 1–4 (ITC 337-TA-1323, 88 FR 52209). Claim 1 is the picture-coding method claim; the substantive elements (mirrored in the decoding/device claims 2–4, which per the district-court record recite an "audio processing unit configured to code/decode audio data" in addition to the video-coding elements) are:

Critical timing premise: the '878 patent claims priority to JP 2002-112665, filed April 15, 2002. The PTAB twice accepted that priority date and found the JP application's written description sufficient — Huawei IPR2018-00658 (institution denied Oct. 9, 2018, as to claim 1) and Dell IPR2022-00629 (institution denied Oct. 4, 2022). In Dell's case the Board expressly held that the asserted references "came in May and August of 2002" — i.e., after the effective filing date — and that "[t]his priority showing is fatal to the Petition's obviousness showing." Dell Techs. Inc. v. VideoLabs, Inc., IPR2022-00629, Paper 10 (PTAB Oct. 4, 2022). Any § 103 combination that depends on post-April-2002 CAVLC documents therefore fails as a matter of law unless the priority claim is defeated — a point I flag in § 5 because it governs which combinations are actually viable.

2. The prior-art universe relevant to the claimed subject matter

A. Pre-April-2002 art (usable under the accepted priority date):

• ISO/IEC 14496-2 (MPEG-4 Visual), Dec. 1999 — expressly admitted in the '878 specification (col. 1, citing § 7.4.1, pp. 119–120). Discloses block-based (4×4/8×8) DCT/transform coding, quantization, zig-zag scanning, run/level representation of coefficients, "code tables" mapping (run,level) → code numbers, and VLC tables mapping code numbers → codewords, with separate intra/inter tables. The specification concedes the single-table-per-mode design it calls deficient.
• ITU-T H.263 (1996; v2 1998) — 3-D (LAST, RUN, LEVEL) VLC coefficient coding, intra/inter VLC tables, and pervasive neighbor-based redundancy exploitation (motion-vector prediction, COD/AC prediction). Dell's petition used H.263 as part of the base combination (Ex. 1005 in IPR2022-00629).
• ITU-T H.324 (1998/2001) — the low-bitrate multimedia terminal framework into which H.263/MPEG-4 coding is embedded (Ex. 1004 in IPR2022-00629). The expert declaration Dell filed for the sibling '238 patent states: "the combination of H.324 and H.263 discloses every limitation of claim 1, except for the 'coefficient n[umber]…'" element — i.e., H.324+H.263 supplies the block transform/VLC environment and the only gap is the context-derived coefficient-count table selection.
• MPEG-2 / H.262, H.261, JPEG — same family of transform-VLC block coders; JPEG in particular codes the DC coefficient of a block as a difference from the previous (left-neighbor) block's DC — an early example of predicting a current block's coefficient statistic from a coded peripheral block.
• VCEG "H.26L" test-model documents and CABAC proposals (2001) — TML long-term models (through TML-9, Dec. 2001) and Marpe/Schwarz/Wiegand context-based arithmetic-coding proposals (early 2001) taught context modeling of transform-coefficient syntax using statistics of neighboring blocks — the same "count of significant coefficients in above/left blocks selects the entropy model" concept, implemented with arithmetic probability models rather than switched VLC tables. (I cannot verify specific VCEG document numbers/dates with confidence from the sources retrieved; the characterization is at the level of the disclosed record.)

B. Post-April-2002 art (closest in substance, but time-barred unless priority fails):

• U.S. 7,099,387 (Bjontegaard) — per the Dell IPR record (Ex. 1006), this patent describes literally the claimed mechanism: "coding logic 124 of Num-Trail encoder block 102 computes an average number of transform coefficients in one or more selected neighboring sample blocks as the context metric (Avg.)" — using the blocks above and to the left — and selects a VLC table by thresholds: "if (0<=Avg<2) select codeword table Num-VLC0…; if (8<=Avg) generate a 6-bit fixed length codeword." It also discloses the decoder is "similarly constituted." This is the closest single reference to claims 1–4.
• JVT-C167 (H.264 Joint Committee Draft, ~May 2002) and JVT-C028 (RealNetworks, July 2002) — the CAVLC design adopted for H.264, combining the number of non-zero coefficients with trailing-ones (Num-Trail) and selecting the coeff_token VLC table from the average non-zero-coefficient count of the above and left 4×4 blocks (the nC context). Per the Amazon briefing, the JVT "considered and rejected a proposal from Nokia—supported by the inventors of the '238 and '878 patents and [Matsushita]" (JVT-C088) and "adopted a proposal from RealNetworks" (JVT-C028) that jointly codes Num-Trail.
• U.S. 6,741,656 (Tanaka) (Ex. 1010 in IPR2022-00629) — an adaptive-VLC reference in Dell's ground; I could not verify its disclosure or filing date from the retrieved sources and so do not rely on it affirmatively.

C. Same-family, non-prior-art context: The '878/'238 specifications descend from the pooled Matsushita "ancestor" patents U.S. 7,095,896 and U.S. 7,308,143 (per the Amazon brief). These share the April 15, 2002 priority date and are not § 103 prior art, but they matter for the obviousness-type-double-patenting history (ITC 337-TA-1323, affirmed Aug. 1, 2023, 88 FR 52209 — claims 1–4 of the '878 held invalid for ODP over the family's own earlier claims).

3. Combination 1 — MPEG-4 (ISO/IEC 14496-2) alone, or with H.263/H.324, plus the known neighbor-prediction toolset (viable under the accepted priority date)

Ground: Claim 1 obvious under § 103 over MPEG-4 in view of H.263/H.324 and the well-known practice of spatial neighbor-context modeling in transform coders (as evidenced by MPEG-4 DC/AC prediction, H.263 motion-vector prediction, and 2001 CABAC context modeling).

• [1a] block transform coding — MPEG-4/H.263 (DCT of 4×4/8×8 blocks, quantization) — admitted prior art.
• [1d] VLC of coefficient syntax using code tables + VLC tables — MPEG-4's two-stage (code number → codeword) run/level VLC, expressly cited in the patent.
• [1b] predicting the current block's non-zero-coefficient count from peripheral coded blocks — MPEG-4 already derives prediction from adjacent reconstructed blocks (DC/AC prediction for intra), and H.26L/CABAC work before April 2002 derived coding contexts for transform-coefficient syntax from the significant-coefficient activity of the above/left neighbors. A POSITA seeking a scalar "activity" predictor for the current block would naturally use the mean of the above and left blocks' non-zero counts — the identical context metric later standardized as nC.
• [1c] selecting the VLC table from that prediction — the specification itself concedes the only reason a single intra table and a single inter table (MPEG-4) underperform is that coefficient-count statistics vary with picture content. The remedy — maintaining several VLC tables and picking one by the predicted count — is the textbook entropy-coding response (multiple Huffman tables selected by context), already deployed in MPEG-4 for the intra/inter dichotomy and in JPEG for DC categories.

Motivation and reasonable expectation of success: The stated problem (MPEG-4's fixed one-table-per-mode VLC yields content-dependent efficiency loss) was a known inefficiency of exactly the MPEG-4 design the specification cites; the field had an established solution vocabulary (neighbor-based prediction + context-selected tables); and the change is confined to the entropy stage, requiring no modification to the transform, quantization, or prediction architecture. A POSITA would combine with a strong expectation of success. This is the most legally viable ground because every reference predates April 15, 2002.

Weakness: MPEG-4/H.263 do not themselves code the total number of non-zero coefficients as a separate syntax element (they use run/level with EOB/last), so element [1b]/[1d] — coding the count itself — depends on importing the count-based context concept from CABAC-era art or on arguing the count is an inherent, obvious reorganization of run/level syntax. This is where the analysis is contestable.

4. Combination 2 — the "CAVLC" ground (Bjontegaard '387 and/or JVT-C167/JVT-C028 + MPEG-4/H.263) — substantively strongest, but time-barred on the current priority date

Ground (if the priority date is defeated): Claim 1 obvious under § 103 over Bjontegaard (U.S. 7,099,387) or JVT-C167/JVT-C028 in view of MPEG-4/H.263.

• Bjontegaard '387 alone maps onto every substantive element: context metric = average of non-zero-coefficient counts of the above/left coded blocks ([1b]); table selection by thresholding the average ([1c]); VLC coding of the count (combined Num-Trail) using the selected table ([1d]); decoder mirror-image ([3]/[4]). The Dell expert record explicitly describes this mapping for the sibling '238 patent.
• JVT-C028 (the CAVLC the JVT actually adopted, July 2002) and JVT-C167 (the Committee Draft) disclose the same nC = (nA + nB)/2 context and coeff_token VLC-table switching, embedded in an H.264 block-transform coder satisfying [1a].

Motivation: This is the least "inventive" combination in substance — the record shows three independent groups (Nokia/Matsushita JVT-C088, RealNetworks JVT-C028, and Bjontegaard's patent) converged on the identical scheme within months of the priority date; the JVT chose between the Nokia and RealNetworks variants on "technical and IPR situation" grounds (Communications Standards Review, vol. 13, no. 25, at 39 (July 24, 2002), quoted in the Amazon brief). Contemporaneous independent convergence of that kind is powerful evidence that the step would have been obvious to a POSITA.

Legal obstacle: Every reference in this ground post-dates April 15, 2002. The PTAB has twice held the JP priority application provides written-description support (IPR2018-00658; IPR2022-00629, where the Board called the priority showing "fatal" to Dell's petition). On the current record, this ground is not citable. It becomes viable only if the '878's priority claim is later defeated — e.g., for claims whose "audio processing unit" limitations Dell argued (unsuccessfully) lacked support in the JP application. This is the single most important caveat in the analysis: the strongest substantive prior art is, on the present procedural record, unavailable as § 103 art.

5. Graham factors and secondary considerations

• Scope/content of prior art: transform-block VLC coding with code tables + VLC tables (MPEG-4, H.263); neighbor-based prediction and context modeling (MPEG-4 AC/DC prediction; 2001 CABAC work); multi-table entropy coding selected by context — all known before April 2002.
• Differences: the combination of (i) coding the count of non-zero coefficients per block and (ii) selecting the VLC table from a neighbor-derived prediction of that count.
• Level of ordinary skill: adopted in IPR2022-00629 — undergraduate degree in EE/CS/CE (or equivalent) plus two or more years in audio/image/video coding, with a working knowledge of H.263 and emerging H.26L/H.264 work (Pet. Ex. 1003 ¶ 24, agreed to by Patent Owner's expert). At that skill level, translating a neighbor-count context into a table-selection index is a routine design choice.
• Secondary considerations: (a) The 2002 JVT record shows simultaneous, independent development of the same CAVLC scheme — evidence cutting against non-obviousness (though those documents are not themselves citable art due to timing); (b) the ITC's ODP holding (claims 1–4 not patentably distinct from the family's earlier claims, 88 FR 52209) does not decide § 103 but corroborates that the claimed advance over the family's own disclosures was incremental; (c) no credible showing of long-felt need or industry praise for this specific count-prediction scheme appears in the retrieved record.

6. Bottom line

1. Viable ground (pre-April-2002 art): Claims 1–4 are most plausibly attacked under § 103 using MPEG-4 (ISO/IEC 14496-2) as primary reference, combined with H.263/H.324 and the pre-2002 neighbor-context modeling known from CABAC/TML work, for the block-transform, code-table/VLC-table, neighbor-prediction, and adaptive-table-selection elements. The motivation is the conceded MPEG-4 single-table inefficiency; the combination is at the level of routine entropy-coding engineering. Its principal weakness is that MPEG-4/H.263 do not themselves code a standalone "number of non-zero coefficients" symbol with context-selected tables.
2. Substantively strongest ground, presently unavailable: Bjontegaard U.S. 7,099,387 and JVT-C167/JVT-C028 (and the Nokia JVT-C088 sibling) disclose the claimed mechanism almost element-for-element, but all post-date the April 15, 2002 priority date that the PTAB upheld in both IPR2018-00658 and IPR2022-00629. This ground would be revived only if a challenger defeats the priority claim (e.g., on written-description grounds for the audio-processing limitations).
3. Framing caution: No reference combination "renders the claims obvious" on the currently accepted priority date with the strength of the CAVLC ground. Any credible § 103 case must either (a) win the priority-date fight to unlock the 2002 CAVLC art, or (b) build the Combination-1 case from MPEG-4/H.263 + pre-2002 context-modeling and argue the missing "count" element is an obvious reorganization of run/level syntax — an argument that is reasonable but will be contested on element [1b]/[1d] mapping.

Key sources: IPR2022-00629, Paper 10 (PTAB Oct. 4, 2022) and exhibit list (H.324, H.263, Bjontegaard 7,099,387, JVT-C167, Tanaka 6,741,656) — https://bannerwitcoff.com/wp-content/uploads/2022/10/IPR2022-00629.pdf; Dell petition declaration re '238 patent — https://ptacts.uspto.gov/.../[1507415](/patent/1507415); VideoLabs v. Amazon claim-construction brief (JVT-C028/JVT-C088 history; '878 claims 1–4 "audio processing unit") — https://ptacts.uspto.gov/.../[1553657](/patent/1553657); ITC final ODP determination, 88 FR 52209 — https://www.federalregister.gov/documents/2023/08/07/2023-16773; Law360/Steine-Kessler summary of the Dell denials — https://www.sternekessler.com/news-insights/news/ptab-rejects-dells-challenges-3-video-tech-patents/.