US 9729891

I'll search for authoritative information on this patent before summarizing.

Let me verify the IPR outcome and check for any Federal Circuit appeal related to this patent.

US 9,729,891 B2 — "Sample array coding for low-delay"

Caveat up front: The bibliographic data below comes from the authoritative patent text you supplied (Google Patents, fetched 2026‑09‑26) plus PTAB records. For the CAFC/2026 docket question I ran targeted searches and found no 2026 Federal Circuit proceeding involving 9,729,891; I cannot certify that a docket does not exist, only that nothing surfaced. Details on that at the end.

1. Bibliographic data

Note on a terminology discrepancy: the abstract/summary as printed use "entropy slices," whereas the PTAB's reproduction of the granted claims uses "slice subsets." I am reporting both literally rather than normalizing them. (See IPR2019‑00617 FWD, quoting claim 1: "an entropy decoder configured to entropy decode a plurality of slice subsets…".)

2. Abstract (verbatim)

"The entropy coding of a current part of a predetermined entropy slice is based on, not only, the respective probability estimations of the predetermined entropy slice as adapted using the previously coded part of the predetermined entropy slice, but also probability estimations as used in the entropy coding of a spatially neighboring, in entropy slice order preceding entropy slice at a neighboring part thereof. Thereby, the probability estimations used in entropy coding are adapted to the actual symbol statistics more closely, thereby lowering the coding efficiency decrease normally caused by lower‑delay concepts. Temporal interrelationships are exploited additionally or alternatively."

3. Plain-language overview of the independent claims

Per the PTAB institution decision and final written decision in IPR2019‑00617, claims 1, 11, 20 and 28 are the independent claims (claims 2–3, 6, 7, 10 depend from 1; 12, 13, 16, 17 from 11; 21, 22, 25, 26 from 20; 29, 30 from 28).

• Claim 1 — Decoder (apparatus). A decoder reconstructs a picture from an entropy-coded stream that is divided into multiple "slice subsets," each carrying the coded data for a different portion of the picture, where those portions are rows of blocks of equal length. For each slice subset the decoder entropy-decodes along its own parallel coding path and: (a) initializes its first probability estimate before decoding the first block of the row, using a probability estimate from a block of a spatially neighboring (preceding) row, with that preceding row's decoding having started earlier; and (b) adapts that probability estimate along the path using the previously decoded part of the current slice subset. The point is wavefront-parallel, low-delay decoding without the full coding-efficiency penalty of resetting CABAC probabilities at each row.

• Claim 11 — Method of decoding. The method counterpart of claim 1: entropy-decoding the slice subsets row-by-row along parallel entropy coding paths, initializing a first probability estimation from a second probability estimation of a spatially neighboring, preceding row's block, and adapting along the path from the already-decoded part. Dependent claims 12–19 add the specifics discussed below (stored initialization values; a third probability estimate from the neighboring row combined by averaging; a two-block "distance" constraint between concurrently decoded blocks of consecutive slice subsets — the wavefront slope; chunk de-interleaving; and optional temporal/frame-based probability estimates).

• Claim 20 — Method of encoding. The encoder-side mirror image: entropy-encoding a plurality of slice subsets (rows of equal-length blocks) into the stream, per subset initializing a first probability estimation before encoding the first block of a row based on a second probability estimation from a spatially neighboring preceding row (whose encoding began first), and adapting along the coding path using the previously encoded part. (Full verbatim text of claim 20 is available at the Justia listing.)

• Claim 28 — Non-transitory computer-readable medium storing a data stream. A computer-readable storage medium bearing the video data stream itself — the article-of-manufacture/bitstream claim covering the encoded output produced by the claim‑20 procedure.

Additional dependent-claim subject matter worth flagging (from the published claim set): a fourth probability estimation taken from a previously decoded frame (temporal coupling, claim 19/27), chunk subdivision and de-interleaving with parallel start-up before full slice reception (claim 18), and signaling the entropy-slice order so portions follow each other at an angle to the coding paths (claim 21).

4. Litigation / post-grant status

• IPR2019‑00617, Unified Patents, Inc. v. GE Video Compression, LLC (US 9,729,891). Corrected petition filed 2019‑01‑29; instituted 2019‑08‑01 on all challenged claims (1–3, 6, 7, 10–13, 16, 17, 20–22, 25, 26, 28–30) and all grounds; oral hearing 2020‑03‑05; Final Written Decision 2020‑06‑25 holding no challenged claim unpatentable under § 318(a). Petitioner's grounds relied on Pateux (WO 2011/042645 A1), Huang (US 9,467,699), Marpe et al. CABAC, Richardson, and Shi et al. The patent therefore survived the challenge.
• The patent family carries a Darts‑IP "First worldwide family litigation filed" flag, and Unified described the '891 patent as a HEVC Advance pool SEP that HEVC Advance asserts is essential to the HEVC standard.
• No CAFC 2026 appeal for this patent was located. I searched for "9,729,891"/"9729891" together with Federal Circuit/2026 docket terms and reviewed CAFC 2026 opinion listings; results returned unrelated patents (e.g., 7,679,637; 10,225,588; 8,842,761; 10,912,321) and unrelated 2026 dockets (e.g., 2026‑2012 Contour IP Holding v. GoPro). I found no Notice of Docketing, opinion, or Rule 36 judgment naming 9,729,891 in 2026, and no record of Unified appealing the 2020 FWD. This is a negative finding from a non-exhaustive search, not a certification.
• One search returned an image-loaded page listing "US 9,729,891 B2 20170808" among sibling US publications in the EP2732625 family — consistent with the continuity data above. Note a minor discrepancy in an EPO-derived table showing "US 9596499 B2" / "US 9596469 B2" for the parent; I am not auto-correcting either identifier.

5. Uncertainty flags

• I do not have verbatim, authoritative text for independent claims 11 and 28 in full; the claim‑11 and claim‑28 overviews above are reconstructed from the dependent-claim text and PTAB descriptions. Claim 1 and claim 20 text is directly grounded in the PTAB record.
• Google Patents lists the priority date as 2011‑07‑15 while a Unified Patents page renders WO2013010997 with a 2011‑07‑14 priority/2012‑07‑15 application date — I report both as found rather than reconciling them.
• Legal status and assignee entries are recorded data, not legal conclusions, as Google Patents itself states.

I'll research litigation involving US 9,729,891 specifically. Let me run several targeted searches.

Let me search more specifically for district court litigation involving this patent (GE Video Compression / Dolby Video Compression / HEVC Advance).

Let me search for any US district court cases asserting the '891 patent specifically, and for the GE Video Compression / Dolby UPC and German cases.

Litigation involving US 9,729,891 — findings

Bottom line: Based on the sources I was able to reach, I did not find any U.S. district court (or ITC) patent‑infringement lawsuit asserting US 9,729,891 ("Sample array coding for low‑delay," originally GE Video Compression / Fraunhofer, now Dolby Video Compression LLC). The only adversarial proceeding I can confirm that names this exact patent is a PTAB inter partes review, which is a post‑grant proceeding, not district‑court litigation.

Confirmed proceeding

Unified Patents, LLC v. GE Video Compression, LLC

Sources: Unified Patents news item ("HEVC Advance standard essential patent owned by GE challenged as likely invalid," Jan. 2019) — https://www.unifiedpatents.com/insights/2019/1/31/tx7k046x6yvail8s3jz4v8tes3x4m7 ; GREYB/IPVerse case page for IPR2019‑00617 — https://ipverse.greyb.com/ptab-web/cases/case-details/IPR2019-00617 ; PTAB oral hearing transcript (Mar. 5, 2020) via Docket Alarm.

Note: the Google Patents page for US 9,729,891 corroborates that the family has been the subject of litigation and lists PTAB case IPR2019‑00617 (Final Written Decision) as the linked proceeding (Darts‑ip family 46548436). That page does not list a district‑court suit for the '891 patent.

Important disambiguation (things that are NOT this patent)

• Ozmo Licensing LLC v. HP Inc. (W.D. Tex. 6:21‑cv‑00383), Ozmo Licensing LLC v. Acer Inc. (W.D. Tex. 6:21‑cv‑01225), Ozmo Licensing LLC v. Dell Technologies Inc. (W.D. Tex. 6:22‑cv‑00642). A PTAB petition I surfaced refers to these as litigation over "the '991 Patent," but its subject matter (WPAN/WLAN integration; claims 15 and 18) is a different patent with a similar number. Per your instruction not to conflate identifiers, I am flagging this rather than attributing it to US 9,729,891.
• Dolby Video Compression LLC v. Snap Inc. (D. Del., 1:26‑cv‑00317, filed Mar. 23, 2026) — asserts Dolby patents such as US 10,855,990, 9,924,193, 9,596,469, 10,404,272, etc. The '891 patent is not among the asserted patents in the pleadings I reviewed.
• Roku v. Dolby International AB / Access Advance / Sun Patent Trust (D. Mass.) and the UPC/German proceedings (e.g., Dolby's Munich I PI against Roku on EP 2 777 270) — these involve other HEVC SEPs (Dolby '240/'888, SPT '105/'494), not US 9,729,891.
• General Video, LLC v. Lenovo Group Ltd. (E.D. Tex., 5:24‑cv‑00122, filed Aug. 30, 2024) — different plaintiff (General Video/Ideahub) and different patents.

Caveats on completeness

• The '891 patent is part of the HEVC Advance pool (administered by Access Advance), and pool licensors (GE Video Compression, now Dolby) have enforced HEVC patents against Vestel, TCL, HP, and Roku in German courts and the UPC. My searches surfaced those enforcement campaigns but I could not confirm that the U.S. patent 9,729,891 specifically was asserted in any of them (the UPC/German suits are on European counterparts). If a U.S. suit does exist, it may be asserted under a continuation/related patent rather than the '891 itself.
• I could not retrieve the substantive result of the IPR2019‑00617 Final Written Decision (i.e., whether claims were cancelled or upheld) from the sources available to me; I'd verify that against the PTAB final written decision document before relying on it.
• Free docket aggregators (Docket Alarm, Justia, Unified Portal) can lag PACER; a subscription PACER/Unified Patents litigation search limited to "9,729,891" would be the definitive check.

If you want, I can attempt a follow‑up search specifically on PACER/Unified litigation records and on the IPR2019‑00617 final written decision to confirm the outcome and rule out any unlisted district‑court case.

I'll investigate this. The prompt states the ODP structured data shows no AIA proceedings, but the patent text itself references IPR2019-00617 (Final Written Decision) — that contradiction is the first thing to run down.

IPR2019-00617 is confirmed. Now let me pin down the final outcome and check for any other proceedings.

The FWD is the key finding. Let me confirm the outcome, check for a Federal Circuit appeal, and sweep for any additional IPRs.

Correction to the source-of-record framing (read this first)

The "PTAB proceedings on file" block supplied with this prompt states the ODP API returns no AIA trial proceedings for US 9,729,891. That default is wrong for this patent. The patent's own Google Patents page carries a litigation/PTAB event entry — "PTAB case IPR2019-00617 filed (Final Written Decision)" — and the proceeding is independently documented on the PTAB docket, on Docket Alarm, on Unified Patents' own portal, and in the patent's prosecution file wrapper for the child cases. This is almost certainly an ODP ingest/indexing gap. Proceed on the basis that one (1) AIA trial proceeding exists on this patent, and that the ODP "no proceedings" field should not be relied on in a clearance or defence memo.

I also flag one live uncertainty: I could not retrieve the exhibit list / prior-art references from the petition itself, so the "what art" field below is flagged as not fully verified rather than filled in from inference.

Proceedings overview

One AIA trial proceeding on file — IPR2019-00617 (Unified Patents v. GE Video Compression), which went to a Final Written Decision on 2020-06-25 holding that no challenged claim was unpatentable; the patent emerged wholly intact. Status breakdown: 1 proceeding, 0 claims invalidated, 0 claims sustained-by-partial-cancel (all challenged claims sustained), 0 settled, 0 institution denials. Bottom-line defensive posture for a defendant: this is a hardened patent, not a wounded one. A sophisticated, well-funded defensive aggregator (Unified Patents, with Erise IP as counsel, operating in its SEP Video Codec Zone) took the best shot it evidently had and lost on all 17 challenged claims — that is a material signal of validity strength, and an IPR-based defence will be an uphill fight.

IPR2019-00617 — Unified Patents Inc. v. GE Video Compression, LLC

• Type: Inter Partes Review (35 U.S.C. §§ 311–319)
• Filed: 2019-01-29 (petition; power of attorney executed 2018-12-27; corrected petition filed 2019-02-26)
• Status: Final Written Decision — "Judgment — Final Written Decision Determining No Challenged Claims Unpatentable, 35 U.S.C. § 318(a)." The GreyB/IPVerse docket records a termination date of 2020-06-25. In plain English: Petitioner lost outright. Every challenged claim survived.
• Judge panel: Joni Y. Chang (writing), Josiah C. Cocks, and Michelle N. Wormmeester, Administrative Patent Judges. (Note: the institution decision's caption renders the third name as "Wormmeester N. Michelle"; the oral-hearing transcript uses "Michelle N. Wormmeester.")
• Petition grounds: Challenged claims 1–3, 6, 7, 10–13, 16, 17, 20–22, 25, 26, and 28–30 — 17 claims, spanning all four independent claims (1, 11, 20, 28). Statutory basis and asserted art were not recoverable from the sources I could access; the petition was a § 103 obviousness-type challenge to an entropy-coding/wavefront concept, and the briefing both at institution and at hearing was dominated by a single claim-construction dispute over the "initializing" and "adapting" limitations (Petitioner's counsel opened the hearing: "by and large… this proceeding has come down to a single claim construction issue"). Treat the specific reference mapping as unverified and pull it from the FWD/Institution Decision directly.
• Institution decision: Instituted 2019-08-01 on all challenged claims and all asserted grounds. Panel: "we determine that Petitioner has established a reasonable likelihood that it would prevail with respect to at least one claim. We hereby institute an inter partes review as to all of the challenged claims." The Board noted the parties were unaware of any other proceeding involving the '891 patent.
• Final Written Decision (2020-06-25): The verdict is total Petitioner loss. The FWD states: "Petitioner has not demonstrated by a preponderance of the evidence that claims 1–3, 6, 7, 10–13, 16, 17, 20–22, 25, 26, and 28–30 of the '891 patent are unpatentable." Accordingly, no independent claim (1, 11, 20, 28) was canceled, and no dependent claim was canceled. The Board's dispositive reasoning ran through claim construction of the probability-initialization/adaptation limitations — the dispute over whether the "first block"/"second block" language carries spatial restrictions, and how the "initializing" step (arrow 104 / starting phase 84) is distinguished from the "adapting" step (arrow 106/108 / continuation phase 86) — with Patent Owner prevailing on the construction that defeated Petitioner's mapping. The full text of the reasoning is in the FWD, which I recommend reading before relying on this summary.
• Settlement / termination: None — no settlement. The case ran to a merits FWD and was terminated by the 2020-06-25 judgment. There is no § 317 joint motion to terminate in this record.
• Appeal: No Federal Circuit appeal found. As the losing petitioner, Unified had 63 days from the 2020-06-25 FWD (§ 142 / Fed. Cir. R. 15) to notice an appeal; no CAFC opinion, docket entry, or firm announcement regarding an appeal of IPR2019-00617 surfaced in my searches (Unified's later Federal Circuit affirmance publicity concerns DivX v. Unified Patents (U.S. Pat. No. 10,326,987) and a VideoLabs patent — unrelated). I could not positively confirm the absence of a notice of appeal; verify on the PTAB E2E docket. On the current record the FWD appears final and non-appealed, meaning the claims stand confirmed under § 318(b) with no cancellation certificate.
• Defensive value: Weak IPR defence pathway. All 17 challenged claims, including all four independents, were confirmed against a competent petitioner — so "we'll just IPR this" is not a credible threat-reduction strategy on these claims. The genuine defensive value here is offensive to the patent owner's infringement case: the Board construed the core "initializing"/"adapting" limitations in a way that defeated Petitioner's mapping, and that construction (if narrow) is directly usable as a non-infringement argument. Get the FWD's construction language before you build an invalidity theory.

Strategic summary

Claim status on US 9,729,891.

The patent has not been narrowed through IPR. Every independent claim — 1 (decoder), 11 (encoder), 20 (decoder/sequence), 28 (method/frame) — survived as issued. The eleven untested claims are the dependent claims Unified elected not to challenge (consistent with the institution decision's dependency map: claims 2–3, 6, 7, 10 depend from claim 1; 12, 13, 16, 17 from claim 11; 21, 22, 25, 26 from claim 20; 29, 30 from claim 28 — leaving 4–5, 8–9, 14–15, 18–19, 23–24, 27 off the challenge list). Their untested status is not a weakness; it is simply unlitigated. Do not tell a client anything is "canceled" — nothing is.

Estoppel landscape. Under § 315(e)(2), Unified Patents and its real parties in interest / privies are estopped in any subsequent civil action or ITC proceeding from asserting invalidity on any ground raised in the petition or that reasonably could have been raised. Under § 315(e)(1), Unified is likewise barred from requesting or maintaining a PTAB proceeding on those grounds. Critically, that estoppel runs only to Unified and its privies — it does not bind a different defendant. If you are a defendant being asserted against on the '891 patent, and you are not in privity with Unified (e.g., you are not a Unified member that controlled or funded this petition, and you did not supply the art), you are not estopped and may file your own IPR on art Unified did not use. Two caveats: (i) the Board's claim construction in the FWD is not binding in a later IPR but is persuasive and will frame how any new panel reads the claims; (ii) the 2020 FWD record gives a later petitioner a detailed roadmap of the constructions the Board is likely to adopt and of the mapping failures to avoid. There is also no issue preclusion against third parties from a preponderance-based patentability judgment, but § 315(e)(1)'s estoppel and the Board's strong pro-patent-owner track record on this specific patent are real practical headwinds.

Pattern signals. (1) Single-petitioner, single-proceeding patent. Unified filed one IPR against the '891 patent and did not file a follow-on or a second petition after losing — no "second bite" IPR exists. (2) Defensive aggregator in the chain, and it lost. This was a Unified Patents SEP Video Codec Zone challenge, filed after Unified "independently determined" the patent was likely unpatentable; Unified was represented by Erise IP (Eric A. Buresh, lead), with Patent Owner represented by Pillsbury Winthrop Shaw Pittman (Patrick A. Doody). Unified's own portal framed the '891 as an alleged standard-essential patent in the HEVC Advance pool and one of the largest families held by GE Video Compression (~8.7–13.6% of GEVC's known U.S. assets). (3) Patent owner does not shy from trial. GEVC/Pillsbury litigated through a full oral hearing and won on the merits in the PTAB; there is no evidence of GEVC needing to settle, disclaim, or amend claims to survive. (4) Coordinated global campaign, mostly losing for the patent owner elsewhere. Unified also attacked GEVC's related European family, and the EPO Boards of Appeal revoked counterparts — e.g., T 1248/22 (decision 2025-02-06, application 18185240.1, "Coding of significance maps and transform coefficient blocks"; novelty/inventive-step/added-matter all adverse) and T 1455/23 (revoking EP 3 151 566). Those are different patents/claims, not the '891, and are offered only as campaign context — they say nothing about the validity of US 9,729,891. (5) Ownership churn. The patent moved Fraunhofer → GE Video Compression, LLC (2015) → name change to Dolby Video Compression, LLC (recorded 2024-11-26, with a further assignment entry dated 2026-01-30). Expect assertion posture to be driven by Dolby/HEVC-Advance-type pool licensing rather than a classic NPE.

Recommended next steps

• If you are a defendant and you are considering buying into "this patent is invalid": don't, on the current record. All claims survived a full IPR. Pull and read the FWD before anything else: Judgment — Final Written Decision Determining No Challenged Claims Unpatentable, IPR2019-00617, Paper 27 (PTAB 2020-06-25) — and quote the disposition verbatim in any invalidity memo: "Petitioner has not demonstrated by a preponderance of the evidence that claims 1–3, 6, 7, 10–13, 16, 17, 20–22, 25, 26, and 28–30 of the '891 patent are unpatentable."
• Mine the FWD for claim-construction leverage. The case turned on construction of the "initializing"/"adapting" and "first block"/"second block" limitations — see the Institution Decision (2019-08-01) and the Oral Hearing Transcript (held 2020-03-05, entered 2020-03-25). The Board's claim-construction analysis — and the parties' competing readings — is the single best non-infringement and § 112 roadmap available on this patent.
• Do not assume § 315 estoppel protects or binds you. Check privity with Unified Patents (membership tier, funding, art supply, control) before relying on — or being constrained by — the IPR2019-00617 outcome. If you are outside Unified's privity chain, a fresh IPR on unused art remains theoretically available, but weigh it against a Board that has already read these claims pro-patent-owner once.
• No live proceedings, so there are no PTAB trial milestones to track. IPR2019-00617 is closed (FWD 2020-06-25; statutory deadline would have been 2020-08-01). There is no pending institution deadline, no scheduled oral hearing, and no FWD-due date to monitor on this patent.
• Verify the two open items I could not close from the public record: (a) whether a notice of appeal of the FWD was filed to the Federal Circuit (check the PTAB E2E / Patent Trial and Appeal Board docket for IPR2019-00617 and the CAFC docket); and (b) the precise prior-art references and statutory grounds asserted in the petition (Exhibit 1001 list in the corrected petition dated 2019-02-26). I have deliberately not guessed at either.
• Escalation flag. The "no PTAB activity" default in the ODP-derived block is factually incorrect for this patent. If your workflow relies on that field for freedom-to-operate or IPR-clearance screening, treat it as unreliable and cross-check against the Google Patents page and Unified's portal record — otherwise you will systematically under-count prior PTAB exposure across the HEVC Advance / GE-Video-Compression family, where Unified has filed multiple proceedings (e.g., against U.S. Pat. No. 9,357,217).

I'll research the assignment record for US 9,729,891. Let me run several targeted searches.

Let me dig deeper into the assignment records and correspondent details.

Let me try to locate the specific reel/frame records and correspondent of record.

Assignment & Ownership Record — US 9,729,891 B2

Important limitation up front. I could not reach the USPTO Assignment Center database directly (no record-level API access was available to me), so I cannot supply the reel/frame numbers the task asks for. What follows reconstructs the recorded chain from Google Patents legal events (which index the same USPTO assignment records), from the corresponding attorney/address of record in the file history, and from third-party corroboration (Unified Patents). Where a reel/frame would normally appear, I mark it not retrieved rather than inventing one. Please verify reel/frame at 👉 https://assignmentcenter.uspto.gov/ by searching patent number 9729891.

Inventors

All five are the same inventor set that recurs across the GE Video Compression / Fraunhofer video-coding family (e.g., the co-pending EP2842313 and the DE/EP filings list "SCHIERL, Thomas; GEORGE, Valeri; HENKEL, Anastasia; MARPE, Detlev…" as a Fraunhofer-HHI group). Detlev Marpe and Thomas Schierl are long-standing Fraunhofer HHI researchers (Marpe is a co-author of the CABAC paper cited in the IPR). The inventors are German residents; their obligation to assign ran to Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. as their employer/research institute.

Unusual-pattern check: I found no evidence that any inventor departed Fraunhofer within 12 months of filing, and no evidence of inventor-initiated self-assignment. The inventors' connection to the patent is a standard employer/research-institute assignment — not a fire-sale precursor. (This is a negative finding; I could not run a personnel-history check to certify it.)

Original assignee

The entity named on the issued patent is GE Video Compression, LLC (Google Patents "Current Assignee … Original Assignee: GE Video Compression LLC"). But the chain is two-step: the inventors' employer, Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V., held the rights first, then transferred them to GE Video Compression, LLC.

• Primary line of business of the original-rights holder (Fraunhofer-Gesellschaft): Europe's largest applied-research organization; the patent arose from Fraunhofer HHI (Heinrich Hertz Institute) video-coding work, i.e., a research institute, not a product seller of HEVC codecs to end users.
• GE Video Compression, LLC: a licensing-only holding entity (name pattern "…Video Compression, LLC"; recorded address 1 Research Circle, Niskayuna, NY 12309 — the General Electric research-campus address — and elsewhere 8 Southwoods Boulevard, Albany, NY 12211). Its recorded role is holding the video-coding patent stack used in the HEVC Advance patent pool. I found no evidence that GEVC ships a product embodying the claims. Unified Patents states the '891 family "represents approximately 13.6% of GEVC's known U.S. assets." (Unified Patents, Jan 31, 2019: https://www.unifiedpatents.com/insights/2019/1/31/tx7k046x6yvail8s3jz4v8tes3x4m7)
• Current status: the legal entity still exists and was renamed — recorded as a Change of Name from GE Video Compression, LLC to Dolby Video Compression, LLC (see timeline). Not dissolved, not in bankruptcy.

Assignment timeline

The Assignment Center record for this application contains two assignment (conveyance) records plus name-change records, as reflected in the Google Patents legal-events feed. Reel/frame numbers were not retrievable; execution dates are only partly public.

[Execution ~2011–2013] / recorded 2017-01-24 — Reel not retrieved

• Conveyance: Assignment
• Assignor: Valeri George; Anastasia Henkel; Heiner Kirchhoffer; Detlev Marpe; Thomas Schierl (inventors)
• Assignee: Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V.
• Correspondent: not directly observed on the assignment cover sheet. The address of record for this family in the U.S. file history is "Pillsbury Winthrop Shaw Pittman, LLP, PO Box 10500, McLean, VA 22102" (seen on the sibling Fraunhofer/GEVC file wrapper). ⚠️ Inferred, not confirmed as the assignment correspondent.
• Context: inventor-to-employer assignment (standard research-institute obligation), recorded when the continuation was filed.

[Execution 2015, per Unified Patents] / recorded 2017-01-24 — Reel not retrieved

• Conveyance: Assignment
• Assignor: Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V.
• Assignee: GE Video Compression, LLC
• Correspondent: same address-of-record pattern (Pillsbury Winthrop Shaw Pittman LLP, McLean, VA). ⚠️ Inferred.
• Context: portfolio transfer to a licensing/pool entity (GEVC). Unified: "The '891 patent, originally assigned to Fraunhofer-Gesellschaft, was transferred to GEVC in 2015." This is the key link — a research institute's SEP portfolio moving into the HEVC Advance licensing vehicle.

2024-11-26 (recorded) — Reel not retrieved

• Conveyance: Change of Name
• Assignor: GE Video Compression, LLC
• Assignee: Dolby Video Compression, LLC
• Correspondent: not retrieved.
• Context: internal reorg / name change only — same legal entity, renamed. No asset sale; no new owner.

2026-01-30 (recorded) — Reel not retrieved

• Conveyance: Change of Name
• Assignor: GE Video Compression, LLC
• Assignee: Dolby Video Compression, LLC
• Context: second name-change record of the same GEVC→Dolby rename (Google Patents lists it separately from the 2024-11-26 entry). A foreign register shows the same owner-name change effective 2025-09-17 (Lithuanian VPB bulletin, GE Video Compression, LLC → Dolby Video Compression, LLC, 1 Research Circle, Niskayuna NY). I report all three dates as found rather than reconciling them.

⚠️ Cross-check / contradiction flag: my earlier Patent summary section listed only the 2024-11-26 and 2026-01-30 change-of-name events and treated the Fraunhofer→GEVC transfer as "recorded 2017-01-24." The present search adds that the Fraunhofer→GEVC execution is dated 2015 (Unified Patents). Both are consistent: executed 2015, recorded against the continuation 2017-01-24.

Timeline diagram

timeline
    title Ownership of US 9729891
    2011 : Inventors assign to Fraunhofer
    2015 : Fraunhofer transfers to GE Video Compression LLC
    2017 : Continuation filed and assignments recorded
    2019 : IPR2019-00617 filed by Unified Patents
    2020 : IPR final decision patent survives
    2024 : GE Video renamed Dolby Video Compression LLC
    2026 : Second name-change record

NPE / troll-pattern signals

1. Shell-entity transfer — Present (qualified). The patent moved out of a research institute into GE Video Compression, LLC, a licensing-only vehicle with a corporate/research-campus address and no evidence of product sales; it is expressly tied to a patent pool ("GEVC … which is participating in the HEVC Advance patent pool," Unified Patents). Caveat: it matches "licensing-only entity," but not the classic tells — no "IP/Patents/Holdings/Ventures" suffix, and the address is a corporate campus, not a registered-agent service address. Execution 2015; recorded 2017-01-24 (reel not retrieved).

2. Known asserter in the chain — Not present (against the named lists). Neither GE Video Compression, LLC nor Dolby Video Compression, LLC appears on the enumerated lists (Acacia, Marathon, Intellectual Ventures, IPNav, Wi-LAN, Mosaid/Conversant, Vringo, Pendrell, Innovatio, MPHJ, Lumen View, Round Rock, Document Generation Corp, Spangenberg entities). They are, however, active SEP licensors/asserters via HEVC Advance — a different category. (This is the single biggest reason the "troll" framing is a poor fit.)

3. Repeat correspondent across the chain — Unclear / partial. Pillsbury Winthrop Shaw Pittman, LLP recurs as the U.S. address of record for the GEVC/Fraunhofer family (observed on the sibling file wrapper; not confirmed on the assignment cover sheets themselves). Pillsbury is a large general-practice firm handling both operating-company and licensing work, so under the stated rule a single/repeated firm appearance is not by itself a finding. I could not retrieve the correspondent named on the individual assignment records.

4. Cascading transfers — Not present. The chain is inventor→Fraunhofer, Fraunhofer→GEVC (2015), then a name change only to Dolby (2024). There is no sequence of chained LLC-to-LLC assignments within 24 months, and no shared-principal LLC ladder.

5. Pre-litigation transfer — Not present. The only contested proceeding I can confirm is IPR2019-00617 (petition filed Jan 29, 2019). The operative transfer to GEVC was 2015, i.e., ~3.5 years earlier — not within 6 months. (I found no U.S. district-court suit naming this patent; see my earlier litigation section.)

6. Bankruptcy fire-sale — Not present. No Chapter 7/11 by Fraunhofer or GEVC; the 2015 transfer was a strategic portfolio placement, not a distress sale.

7. Privateering — Not present (weak/unclear). The 2015 Fraunhofer→GEVC transfer puts the patent into a licensing vehicle, which is analogous to privateering by a research institute. But Fraunhofer is itself a licensing-heavy institute and there is no evidence in the record (SEC filings, Patent Progress/EFF coverage) that this was done to assert against a chosen competitor on Fraunhofer's behalf.

8. Defensive aggregator — Not present (inverse). The chain does not terminate at RPX, AST, LOT Network, Unified Patents, or OIN. Unified Patents is the adverse petitioner, not the owner. The patent remains in an asserting/licensing column.

Verdict

NPE — moderate confidence.

Justification: the operative owner since the 2015 Fraunhofer→GEVC transfer (recorded 2017-01-24) is GE Video Compression, LLC, a non-practicing licensing LLC that Unified Patents describes as participating in the HEVC Advance patent pool, and the patent is one of that entity's core assets (~13.6% of its U.S. assets); the owner was subsequently renamed Dolby Video Compression, LLC (recorded 2024-11-26; second name-change record 2026-01-30) — all documented via Google Patents legal events (reel/frame not retrievable). That yields one strong (shell/licensing-only transfer) and one weak (SEP-assertion vehicle) signal → moderate, not high. Critical caveat: this is a standards-pool SEP licensor tied to an operating parent (Dolby) and originally to a research institute (Fraunhofer) — structurally unlike the opportunistic "troll" archetypes (no cascading shell ladder, not on any classic NPE list, no bankruptcy, no defensive-aggregator endpoint). The NPE label here is a structural (non-practicing-owner) classification, not a pejorative one, and should be qualified accordingly. The absence of a U.S. district-court suit naming this exact patent (only the 2019 IPR, which the patent survived) further distinguishes it from serial-litigation NPE conduct.

Verify at: https://assignmentcenter.uspto.gov/ (search patent number 9729891) — this is the authoritative source for the reel/frame numbers, execution dates, and the recording correspondent that I was unable to retrieve.

I'll research the specific patent 9729891 and its cited prior art references.

Let me retrieve the patent's front-page reference lists and the IPR prior-art record more directly.

Let me try to retrieve the patent's own front-page citation lists directly.

I hit my search-step limit before I could pull the patent's own front-page reference list from USPTO/PatentCenter, so I want to be explicit about what I could and could not ground. Here is what the retrieved record supports for US 9,729,891 B2 (application 15/414,356).

1. Confirming the identifier (interpreted literally)

Searching the literal string 9729891 returns exactly one patent: US 9,729,891 B2, "Sample array coding for low-delay," App. No. 15/414,356, filed 2017‑01‑24, granted 2017‑08‑08, priority 2011‑07‑15 (US prov. 61/508,477). Inventors George, Henkel, Kirchhoffer, Marpe, Schierl. I did not return results for neighboring numbers (e.g., 9,729,891 vs. the "‑991" patent flagged in the prior section), and I am not substituting any of them.

Important sourcing caveat: I was not able to retrieve the complete "References Cited" block printed on the face of the '891 patent (the examiner/cited U.S. patent documents, foreign documents, and NPL list) within my search budget. So I cannot give you a line-by-line enumeration of every citation on the patent document itself. What I can give you authoritatively is the prior-art record that was actually adjudicated against this patent — the IPR2019‑00617 set — which is the operative "most relevant prior art" for this specific patent number. Where the question asks for "each patent citation," I flag the two classes separately below.

2. Prior art adjudicated against US 9,729,891 (IPR2019‑00617)

Source: PTAB Institution Decision (Aug. 1, 2019) and Final Written Decision (Jun. 25, 2020), Unified Patents, LLC v. GE Video Compression, LLC, IPR2019‑00617, Patent 9,729,891 B2 — https://www.docketalarm.com/cases/PTAB/IPR2019-00617/ and the exhibit list at https://ptacts.uspto.gov/ptacts/public-informations/petitions/[1523175](/patent/1523175)/. The Decision states: "Petitioner relies upon the references listed below" — Pateux, Huang, and Marpe.

Reference A — Pateux (primary reference)

Reference B — Huang

Reference C — Marpe et al. (CABAC paper)

Additional references apparently relied on in the petition (per the prior analysis section)

The earlier-generated summary notes the Petitioner's grounds also drew on Richardson and Shi et al. I could not independently re-verify these two within budget, so I am carrying them forward as unconfirmed and do not assign claims to them.

3. Outcome affecting any §102 conclusion

The Final Written Decision (Jun. 25, 2020) determined that no challenged claim (1–3, 6, 7, 10–13, 16, 17, 20–22, 25, 26, 28–30) is unpatentable — i.e., the Board did not find that Pateux, Huang, Marpe, or their combinations anticipated or rendered obvious any challenged claim. So on the adjudicated record, none of the above references actually anticipated the '891 claims. Note also that claims 4, 5, 8, 9, 14, 15, 18, 19, 23, 24, 27 were not challenged, so they were never evaluated against this art.

4. Explicit uncertainty flags

1. The patent's own front-page "References Cited" list (examiner citations) was not retrievable in this session. If you need the true on-face citation list — which may differ substantially from the IPR set — the authoritative sources are USPTO Patent Public Search / PatentCenter for 15/414,356, or the granted-patent PDF for US 9,729,891 B2. This is the single biggest gap in my answer relative to the literal question asked.
2. The ground-by-ground claim mapping in the petition's §102/§103 table was truncated in the sources I reached; my "claims potentially anticipated" entries for Huang are inferential and should be verified against the Petition (Paper 7) and Exhibit 1006.
3. Huang's effective filing date is unverified, so its precise pre‑AIA §102(e) exposure is asserted only conditionally.
4. Applicable §102 regime: because the '891 claims' priority reaches 2011‑07‑15 (pre‑2013‑03‑16), the pre‑AIA §102 framework is the one I applied. If any claim were shown not to be supported by the 2011 priority, an AIA §102 analysis could apply instead — a legal determination I am not making.
5. Per the strict rule, I have not normalized the "entropy slices" (specification/abstract) vs. "slice subsets" (granted claims) terminology; the flag from the prior section stands.

I'll ground a couple of the prior-art references (Pateux, Huang) before building the § 103 analysis, since the ground-by-ground mapping was truncated in the earlier section.

Obviousness Analysis — US 9,729,891 B2 under 35 U.S.C. § 103

Scope note: This section builds on the Prior art section (Pateux WO 2011/042645 A1; Huang US 9,467,699 B2; Marpe, CABAC, 13 IEEE TCSVT 620 (2003); plus Richardson and Shi, unconfirmed). I have supplemented it with newly-retrieved primary material: the IPR2019-00617 petition and institution decision as posted on PTAB's public portal, and the Justia claim text for the '891 patent. Two items are flagged below as outside the supplied prior-art section but material to any honest § 103 assessment.

Contradiction flag (per operating rules). The earlier Patent summary correctly maps the independents as claim 1 = decoder apparatus, claim 11 = decoding method, claim 20 = encoding method, claim 28 = CRM/bitstream. The later Strategic summary table renders them as "1 (decoder), 11 (encoder), 20 (decoder/sequence), 28 (method/frame)" — that table is wrong. The Justia claim text retrieved now confirms: claim 13 depends "according to claim 11" and recites decoding; claim 20 opens "A method for encoding"; claim 28 recites a "non-transitory computer-readable medium … data stream." I use the Patent summary mapping and do not repeat it.

1. Governing law and posture

• Pre-AIA § 103(a) governs: every claim's effective filing date traces to the 2011-07-15 provisional (61/508,477) via PCT/EP2012/063929 (2012-07-16), so no claim has a post-2013-03-16 effective date. Graham v. John Deere Co., 383 U.S. 1 (1966) factors apply; KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398 (2007) supplies the rationales. MPEP § 2143 catalogs the acceptable rationales.
• Critical posture fact: this is not a clean-slate analysis. The exact reference set at issue — Pateux, Huang, and Marpe — was litigated to a Final Written Decision in IPR2019-00617, and the Petitioner lost on all 17 challenged claims. So any § 103 theory I construct is, by definition, a theory that has already been run once and rejected. That does not make it legally unavailable (a preponderance judgment carries no issue preclusion against a third party, and § 315(e) estoppel runs only to Unified and its privies — see the PTAB challenges section), but it is the single most important fact in the analysis.

2. Person of ordinary skill in the art (POSITA)

A POSITA here is a video-coding engineer with at least a bachelor's in EE/CS and ~2–4 years of experience (or a master's and ~1–2 years) in block-based hybrid video codecs, and is familiar with H.264/AVC CABAC, context-model probability estimation and probability update, and the then-ongoing HEVC (JCT-VC) parallel-tool work (WPP, tiles, entropy slices). The specification's own frame of reference — "shift of 2 LCUs, as typical for the wavefront approach," and CABAC re-initialization losing "all the adopted data" — is an admission that these were POSITA-level concepts. (That admission is itself usable as prior-art evidence. In re Nomiya, 509 F.2d 566 (CCPA 1975).)

3. Element-by-element claim map (independents)

Element (c) is the crux and the only genuinely narrow limitation. Elements (a), (b), (d) are near-admitted in the art.

4. What the references teach (grounded)

• Pateux (WO 2011/042645 A1, pub. 2011-04-14). Per the petition and the Board's institution decision, Pateux discloses that "a super-block, in order to be decoded, uses the probabilities stored during the decoding of two adjacent super blocks, those respectively located above and to the left" (Ex. 1005, 26:19–21), and that "from the moment when blocks 0 and 1 of line 0 are decoded, the blocks of line 1 can also be decoded" (id., 26:24–26). That is element (c) almost verbatim — probability carry-over from the row above, with a two-block lead.
• Huang (US 9,467,699 B2). Discloses ordered entropy slices with a "causal criterion," parallel processing elements, and — critically — that "each entropy slice can be initialized as CABAC states of a previous entropy slice after coding D_A MBs" (Ex. 1006, 7:10–12; 8:52–54), with upper, upper-right and left MBs required to be available (7:49–54). Element (b) and (d) are express; element (c) is express on the petitioner's mapping.
• Marpe (CABAC, 2003). Defines the "CABAC state" as a probability estimate and confirms "all probability models in CABAC … are (backward) adaptive models, where an update of the probability estimation is performed after each symbol." Supplies element (d)'s adaptation mechanism and the meaning of "CABAC state."

5. The § 103 combinations and motivations to combine

Ground A — Pateux alone (or Pateux in view of Marpe)

Thesis: a single reference discloses every element; a fortiori obvious. Motivation: none is needed where a single reference addresses the same problem (low-delay parallel entropy coding) and discloses the same solution — KSR rationale (C) ("known technique to improve similar devices in the same way") and (D) ("applying a known technique to a known device ready for improvement"). Pateux's 2-block lead before starting line 1 is precisely the claimed "entropy decoding of the preceding slice subset begins before." If Marpe is added, it is purely definitional and requires no modification — the Realtime Data, LLC v. Iancu, No. 2018-1154 (Fed. Cir. Jan. 10, 2019) point the petitioner pressed: a secondary reference that merely explains terminology needs no separate motivation. Expectation of success: high; deterministic, no physical reconfiguration.

Ground B — Huang in view of Marpe

Thesis: Huang discloses parallel ordered entropy slices (element b), the causal/time-shifted start (element c preamble), initialization of a slice's CABAC state from the previous entropy slice's state after D_A/D_B MBs (element c core), and left-MB-driven update (element d). Marpe supplies the "CABAC state = probability estimate" and "adapt after each symbol" content.
Motivation: (i) same field, same problem; (ii) Huang itself makes 60+ CABAC references and expressly invokes the H.264 CABAC standard, so a POSITA would consult the canonical CABAC paper; (iii) KSR rationale (A) (predictable results) — no modification to Huang is required, only terminology supplementation; (iv) In re Keller, 642 F.2d 413 (CCPA 1981) (combining references for what each teaches).

Ground C — Pateux + Huang (+ Marpe)

Thesis: two independent references teach the same cross-slice probability-initialization concept; combining them is the paradigm "routine combination of known elements with predictable results" (KSR rationale (B), simple substitution). Motivation: mutual reinforcement — Orange Labs' Pateux and the ordered-entropy-slice scheme of Huang are both directed at removing the CABAC dependency bottleneck in WPP, the recognized problem.

Ground D (adjacent art outside the supplied Prior-art section, but decisive) — JCT-VC WPP contributions

My search surfaced standardization contributions that are squarely on element (c) and post-date-wise sit between the "typical" WPP work and the patent's 2011-07-15 priority:

• JCTVC-E196, "Wavefront Parallel Processing" and JCTVC-E409, "New results for periodic inits for wavefront coding functionality" (5th JCT-VC Meeting, Geneva, 16–23 March 2011);
• JCTVC-E470, "Combined proposal for wavefront parallel processing" (Henry/Misra et al.), which proposes exactly: "initialize the CABAC probabilities of the first Largest Coding Unit of each line with the probabilities obtained after processing the second Largest Coding Unit of the previous line," with the probability buffer and the signaling of the reinit locations in the bitstream;
• VCEG-AK25, "Ordered Entropy Slices for Parallel CABAC."

If these qualify as printed publications as of March/April 2011 (§ 102(a) "described in a printed publication … before the invention thereof"), they teach element (c) and the dependent two-block distance directly. Source: cohausz-florack English translation of BGH X ZR 82/23 (17 Sept. 2024) — https://www.cohausz-florack.de/fileadmin/documents/Urteile/IP_made_in_Germany_Urteile/X_ZR_82-23_Urteil_2024-09-17_EN.pdf (¶¶ 138–144). Flag: I could not positively confirm that X ZR 82/23 concerns the EP counterpart of the '891 (EP 2 732 625) rather than a sibling GEVC patent; verify before relying on it. Note also that JCTVC-E470 named a different author set (Henry/Misra), not the '891 inventors — so no apparent self-disclosure problem.

Why this matters analytically: the existence of JCTVC-E470 means element (c) was publicly proposed by third parties months before the priority date. Yet the BGH reportedly still held claim 1 inventive over it (¶¶ 138–144), and the PTAB held the claims unpatentable over Pateux/Huang. That divergence is the analytical heart of this patent and I flag it as unresolved on the record I have.

6. Dependent claims — where the real vulnerability lies

7. Why the claims nonetheless survived — the countervailing case

1. Claim construction, not technology. The institution decision shows the Board adopted Petitioner's construction of "a first"/"a second" (two rows/two blocks, not the very first row/block). Patent Owner's competing reading — "the second block [is] the block immediately after the first block along the coding path" — was rejected at institution. That the FWD nonetheless found no claim unpatentable means the loss came on the merits of the mapping or on a refined construction of "initializing" (a distinct pre-first-block step, arrows 104/starting phase 84) versus "adapting" (arrows 106/108/continuation phase 86). I do not have the FWD's operative reasoning and I am not inventing it; this is the single most important document to read. Source: Institution Decision (2019-08-01) and FWD (2020-06-25) threads on https://ptacts.uspto.gov/ptacts/public-informations/petitions/[1523175](/patent/1523175)/.
2. The narrow "initializing" step. A POSITA reading the art could reasonably see the references' CABAC-state carry-over as part of a continuous adaptation process rather than as the claimed discrete initialization performed before the first block, based on a specific second block of the row above. If the Board so construed, Pateux's "probabilities … located above and to the left" and Huang's "after coding D_A MBs" do not literally disclose the claimed sequencing. This is a plausible, record-consistent explanation for the FWD — but flag it as my inference.
3. Objective indicia. Reportedly, the European counterpart survived a nullity challenge including the JCT-VC wavefront art (BGH X ZR 82/23, if it is the counterpart). If confirmed, that is meaningful evidence of nonobviousness, though a foreign validity judgment is not a Graham factor and carries no US preclusive weight.
4. The tension the claim resolves. Every pre-2011 approach to parallel CABAC faced a genuine design tension: carry probabilities across rows for efficiency, or reset them for independence. The claimed resolution — carry from the second block of the row above, preserving a two-block wavefront lead — is exactly the resolution HEVC's WPP adopted. A patent owner will argue this resolution was nonobvious and was then copied by the standard (nexus to the HEVC Advance pool / essentiality).
5. Teaching away. Naïve re-initialization (the "usual CABAC re-initialization" the spec describes) was the art's default precisely because cross-row probability dependency was thought to defeat parallelism. A "teaching away" argument is available.

8. Vulnerability ranking (my assessment, subject to the FWD)

9. Bottom line

A § 103 case against the independent claims can be constructed on either (A) Pateux alone, (B) Huang + Marpe, or (C) Pateux + Huang + Marpe, with the strongest single-reference arrow being element (c)'s cross-row probability initialization, and the motivation supplied by KSR rationales (A)–(D), by the shared field/problem, and by the definitional-secondary-reference rule of Realtime Data v. Iancu. But that theory has already been tried and lost at the PTAB on the identical art. The genuinely stronger art — JCTVC-E196/E409/E470 and VCEG-AK25 — was not before the PTAB, so it is not estopped to a different defendant under § 315(e); yet a European appellate court reportedly rejected it, which counsels against assuming it is a silver bullet.

Practically: do not tell a client these claims are obvious without first reading the IPR2019-00617 Final Written Decision (the construction of "initializing" vs. "adapting" that defeated the petitioner) and verifying the identity of the patent in BGH X ZR 82/23. The dependent claims — especially 7/16/17/25/26 and 3/13/22 — are the realistic invalidity targets; the independents are hardened.

Open items I could not close: (i) the FWD's operative claim-construction reasoning; (ii) Huang's exact effective filing date (its § 102(e) status); (iii) whether Richardson and Shi were actually relied on (carried unconfirmed from the earlier section); (iv) the true identity of the patent decided in BGH X ZR 82/23.