US 9953240

I'll verify against live sources and check for any 2026 CAFC docket activity involving this patent.

Let me check the litigation history and any appellate activity more specifically.

US Patent 9,953,240 — Verification Summary

Note on sourcing: I could not access a live USPTO PatentCenter/Public PAIR or CAFC docket system directly. The findings below are grounded in the authoritative full patent text supplied (Google Patents mirror of the USPTO record, fetched 2026-09-30) plus live web results for litigation. Where I could not verify something, I say so. Also: your prompt header says "April 26, 2026," while the system date given to me is 2026-09-30; I flag this so you can reconcile the as-of date.

Bibliographic Data

Abstract (as printed): "An image processing system includes: a first identification unit that identifies a static area from an input image captured at each of a plurality of time points; an image generation unit that generates a first image by using the static areas of respective input images captured in a first time span from a processing time point and generates a second image by using the static areas of respective images captured in a second time span from the processing time point; and a second identification unit that compares the first image and the second image and identifies an area having a difference."

Note the abstract describes only two images, while the granted independent claims require three. The spec's "Supplementary Notes" likewise recite two images. This mismatch is worth noting if you are assessing claim scope versus the disclosure.

Plain-Language Overview of the Independent Claims

Claim 1 — System (memory + at least one processor), four functional requirements:

1. Identify "static areas" in input images captured at multiple time points — importantly, "static" is defined broadly to include areas where images show motion smaller than a threshold value (not merely zero motion).
2. Generate three images — a first, second, and third image — each built from the static areas of input images captured within first, second, and third time spans measured back from a processing time point. (These are the "time windows" / background images of the spec.)
3. Compare the first image and second image and identify an area that has a difference (i.e., the candidate static-object area).
4. Classify static objects by length of static time period, based on comparing all three images (first, second, and third).

Claim 5 — Method: the same four steps, framed as an image-processing method. (Claim 5 contains a drafting inconsistency — "comparing the first image and the second image and identify an area having a difference" — reproduced literally as written.)

Claim 6 — Non-transitory computer-readable recording medium storing instructions that cause a computer to perform the same four-step method (same phrasing inconsistency as claim 5).

Dependent claims (2, 3, 4 depend from 1; 7, 8, 9 depend from 5):

• 2 / 7 — the second time span is longer than the first time span.
• 3 / 8 — generate a plurality of images (first, second, third) across a plurality of corresponding time spans, and compare the plurality of images.
• 4 / 9 — the images are generated by calculating a mode, median, or average of pixel values of the static areas.

Practical gist: split each frame into moving (foreground) vs. static (background) regions; build several time-window "background" images purely from the static pixels; diff them; and use the differences across multiple windows to detect and rank/categorize objects that have been stationary for different durations. The claimed advance over the cited prior art (see below) is the foreground/background pre-separation before time-averaging, plus multi-window classification.

Litigation / Docket Check

CAFC 2026: I found no Federal Circuit docket or 2026 appellate activity naming US 9,953,240 or application 14/894,143. Searches returned no matching CAFC entry — so I have no authoritative confirmation of a 2026 appeal. Treat "no CAFC 2026 docket found" as a null result, not proof of absence.

District court litigation (from the patent record's litigation links, not independently PACER-verified):

• 2:24-cv-00720, E.D. Tex. — NEC Corporation v. Anker Innovations Technology Co., Ltd. et al, filed 2024-09-03 (confirmed via RPX litigation database).
• 2:24-cv-00881, E.D. Tex.
• 1:24-cv-01870, E.D. Va.

The "First worldwide family litigation filed" flag on the record points to family ID 51988480. I did not verify whether US 9,953,240 itself is the asserted patent in each of those cases.

Prior Art Reference (key)

• PTL 1 = JP 5058010 B2 (Mitsubishi Electric Research Laboratories; priority 2007-04-05) — corresponds to US 2008/0247599 A1, Porikli, "Method for Detecting Objects Left-Behind in a Scene." This is the multi-time-scale long-term/short-term background model the patent distinguishes over (noting its weakness when a scene has constant change, e.g., continuous pedestrian flow).

Uncertainty Flags

• The certificate of correction (2025-03-18) — I could not retrieve its contents; given the granted claims recite a third image and a classification step absent from the spec's Supplementary Notes, it is possible (not confirmed) the correction affected claim text.
• Whether the granted claim set differs from the pre-grant US20160125268A1 claims — not verified.
• No CAFC 2026 activity confirmed.

I'll search for litigation involving US patent 9953240 using the patent litigation sources.

Let me search the specific litigation sources referenced by the patent.

I found one relevant case. Let me dig deeper into the specific cases referenced by the patent.

Litigation Involving U.S. Patent No. 9,953,240 (US 9,953,240 B2)

Patent identification (literal): U.S. Patent No. 9,953,240 B2 (application US 14/894,143), titled "Image processing system, image processing method, and recording medium for detecting a static object." Original/current assignee: NEC Corporation. Priority date 2013-05-31; granted 2018-04-24.

I located litigation data from the Google Patents litigation docket links for this patent's family (Family ID 51988480) and corroborating litigation databases. Note: the sources I could reach did not allow me to fully pull PACER/CAFC dockets for every case, so some details are flagged as unconfirmed.

Confirmed / reported cases

1. Power Mobile Life, LLC (et al.) v. NEC Corporation

• Court / Jurisdiction: U.S. District Court for the Eastern District of Virginia (E.D. Va.), Alexandria Division
• Case number: 1:24-cv-01870
• Filed: 10/24/2024
• Plaintiff(s): Power Mobile Life, LLC (listed "et al.")
• Defendant(s): NEC Corporation
• Nature of suit: 830 – Patent; cause of action shown as "Patent Non-Infringement – Declaratory Judgment" (i.e., a declaratory-judgment action rather than a straight infringement suit).
• Outcome / status: Listed as Closed (per the case tracker pages below).
• Sources: Google Patents family litigation link (E.D. Va. case 1:24-cv-01870); RPX Empower docket ("Power Mobile Life, LLC et al v. NEC Corporation," E.D. Va., filed 10/24/2024, docket updated daily); Ex Parte case page ("Power Mobile Life LLC v. NEC Corp," filed 10/24/24, status Closed).

2. Eastern District of Texas – Case 2:24-cv-00720

• Court / Jurisdiction: U.S. District Court for the Eastern District of Texas (Marshall Division — case prefix "2:")
• Case number: 2:24-cv-00720
• Plaintiff(s): Not confirmed from the sources I retrieved
• Defendant(s): Not confirmed from the sources I retrieved
• Filing date: Not confirmed (2024 docket series)
• Outcome / status: Not confirmed
• Source: Listed on the Google Patents US9953240B2 page as "US case filed in Texas Eastern District Court" with a Unified Patents litigation link: https://portal.unifiedpatents.com/litigation/Texas%20Eastern%20District%20Court/case/2%3A24-cv-00720

3. Eastern District of Texas – Case 2:24-cv-00881

• Court / Jurisdiction: U.S. District Court for the Eastern District of Texas (Marshall Division — case prefix "2:")
• Case number: 2:24-cv-00881
• Plaintiff(s): Not confirmed from the sources I retrieved
• Defendant(s): Not confirmed from the sources I retrieved
• Filing date: Not confirmed (2024 docket series)
• Outcome / status: Not confirmed
• Source: Listed on the Google Patents US9953240B2 page as "US case filed in Texas Eastern District Court" with a Unified Patents litigation link: https://portal.unifiedpatents.com/litigation/Texas%20Eastern%20District%20Court/case/2%3A24-cv-00881

Family-level litigation flag

The Google Patents record also carries a "Family has litigation – First worldwide family litigation filed" indicator pointing to the Darts-IP global patent litigation dataset for Family ID 51988480, confirming that litigation exists at the family level.

Caveats and limitations

• Because US 9,953,240 is part of the NEC "static object detection" family (ID 51988480), the Texas and Virginia 2024 cases are associated at the patent-family/litigation-link level. I could not independently verify from the retrieved materials that US 9,953,240 specifically is an asserted patent in each of the Texas cases — please confirm the asserted patents against the actual complaints on PACER.
• I was unable to confirm the party names, filing dates, or status for the two E.D. Tex. cases (2:24-cv-00720 and 2:24-cv-00881) from the sources reached. No outcome information is available for them here.
• Do not rely on the party labels alone without PACER verification. In case 1:24-cv-01870, one tracker shows Power Mobile Life, LLC as the plaintiff/DJ-seeker against NEC, while the case is styled as a declaratory judgment of non-infringement.
• A separate 2024 NEC case, NEC Corporation v. Becker Professional Development Corporation (N.D. Tex. 4:24-cv-00647), appeared in searches but concerns multimedia content delivery patents — it is not this patent and is excluded.
• I found no CAFC appeal and no IPR/PGR specifically identified as involving US 9,953,240 in the sources retrieved.

If you need the Texas case details nailed down, the authoritative next steps are to pull dockets 2:24-cv-00720 and 2:24-cv-00881 directly from PACER and the Unified Patents portal links above.

I'll start by checking the structured data (no AIA trials returned) and then probing the web for any PTAB or litigation activity.

No PTAB hits yet. Let me probe harder for any AIA trial tied to this patent or the NEC v. Anker campaign.

Proceedings overview

Total AIA trial proceedings on US 9,953,240: 0. Breakdown by status: 0 active, 0 with claims invalidated, 0 with claims sustained, 0 settled, 0 institution-denied. The USPTO Open Data Portal (the canonical source in the structured "PTAB proceedings on file" block) returns no AIA trials, and my independent web searches surfaced none either — no IPR, PGR, or CBM has ever been filed against this patent. The bottom-line defensive posture is therefore "untested at the PTAB": unlike a patent that has survived co-pending invalidity challenges, '240 presents no PTAB record, no claim-level adjudication, and no Federal Circuit appeal. That cuts both ways — there is no canceled claim to hand you a free win, but there is also no hardening precedent, and all nine claims remain live and presumptively valid. The most useful defensive signal is not a PTAB outcome at all; it is a litigation fact (below): NEC voluntarily dropped '240 from its lead case after a § 101 challenge.

Because there are zero proceedings, there are no per-proceeding entries to report. I will not invent docket numbers or outcomes.

Context that matters more than a PTAB record (litigation, not AIA trials)

The patent is being actively asserted, which is the reason a defendant would want PTAB data in the first place:

• NEC Corp. v. Anker Innovations Technology Co., Ltd. et al., No. 2:24-cv-00720-JRG (E.D. Tex.) — Judge Rodney Gilstrap. NEC asserted six object-detection/image-processing patents against Anker's smart-doorbell products. Anker moved for judgment on the pleadings under Rule 12(c) that four of them — including '240 — were patent-ineligible under 35 U.S.C. § 101. NEC then dropped the '240 Patent from the case (Dkt. No. 98 at 1–2), and the court DENIED AS MOOT Anker's motion as to '240 in its 2026-06-04 memorandum opinion and order. (CourtListener: https://www.courtlistener.com/opinion/[10894127](/patent/10894127)/nec-corporation-v-anker-innovations-technology-co-ltd-and-anker/ ; GovInfo: https://www.govinfo.gov/content/pkg/USCOURTS-txed-2_24-cv-00720/pdf/USCOURTS-txed-2_24-cv-00720-0.pdf)
• Other matters listed in the structured data: No. 2:24-cv-00881 (E.D. Tex.) and No. 1:24-cv-01870 (E.D. Va.).
• A second NEC v. Anker suit was filed 2026-09-18, No. 2:26-cv-00847 (E.D. Tex., Marshall Division), asserting a different set of six patents (US 10,223,619; 10,552,713; 10,706,336; 11,144,776; 12,423,985; 12,641,315). '240 is not among them.
• Prosecution/post-grant housekeeping: a certificate of correction issued 2025-03-18; maintenance fees paid through the 8th year on 2025-10-16.

Strategic summary

Claim status. All claims of '240 are UNTESTED — none has ever been canceled, confirmed, or construed by the Board. The patent has 9 claims with three independents: claim 1 (image processing system), claim 5 (image processing method), and claim 6 (non-transitory computer readable recording medium). Claims 2–4 depend from claim 1; claims 7–9 depend from claim 5. Because no IPR/PGR exists, there is no "claims 1–5 canceled" narrative and no narrow survivor set to work with — the full claim set is in play.

Estoppel landscape. With no institution, no 35 U.S.C. § 315(e)(2) estoppel has attached to anyone — there is no "grounds raised or reasonably could have been raised" bar from a prior petitioner. That is generally favorable for a new defendant. The operative statutory constraint is instead § 315(b): a petition must be filed within one year of service of a complaint alleging infringement. Anker was sued in 2024, so its one-year window has almost certainly closed, and any privies are equally time-barred. A new defendant (e.g., a party later served in the 2026 campaign or a fresh target) would have a clean one-year runway and a clean § 315(e)(2) slate. There is also no prior PTAB record to trigger § 325(d) "same or substantially the same art" discretionary denial — a new petitioner gets a first-bite-at-the-apple posture on art the Examiner never saw.

Pattern signals. No serial-filing petitioner, no defensive aggregator (no Unified Patents IPR in the chain), and no PTAB appeal activity — because there has been no PTAB activity at all. The pattern here is a patent-owner enforcement campaign run through district court, not through the Board: NEC is a serial filer against Anker (2024 and again 2026), and it has shown a willingness to drop a patent rather than litigate a § 101 eligibility motion to decision (the '240 move in the 2024 case). Note the patent family is broader than '240 — related patents US 10,776,931 and US 9,934,576 appear in the "Similar Documents"/family citations and could attract separate challenges; keep an eye out for IPRs aimed at those siblings rather than at '240.

Recommended next steps

• Do not expect PTAB ammunition. There is no Final Written Decision to cite and no invalidated claim to lean on. Verify the absence yourself before relying on it: https://ptacts.uspto.gov/ and the PTAB E2E / USPTO PatentCenter record for US 14/894,143. If an IPR surfaces that the ODP ingest has not yet indexed, treat this section as superseded and re-run the analysis.
• If you were served within the last year, an IPR is available — and cheap insurance. You have no § 315(e)(2) estoppel to worry about, and § 325(d) likely favors you because this patent has never been before the Board. Focus § 102/§ 103 grounds on the art cited on the face of the patent (e.g., US 2007/0280540 (NEC), US 2008/0247599 (Porikli), US 2010/0231721, and JP 5058010 — the PTL 1 reference the specification itself distinguishes), and on the multi-time-window background-modeling concept. Watch the one-year § 315(b) clock from your service date.
• If you were served more than a year ago (e.g., you are Anker or a privy), the IPR door is closed. Pivot the defense to the district-court track. The most promising lever is § 101: NEC's decision to drop '240 from the 2024 case after Anker's Rule 12(c) eligibility motion is a strong, citable signal that the claims are vulnerable on the Alice/§ 101 axis. Pair that with a § 112 enablement/written-description attack on the "static area" / "differs by a threshold value" limitations, and standard § 103 invalidity contentions in the district court, none of which are foreclosed by any IPR estoppel.
• Watch the 2026-09-18 suit (2:26-cv-00847). It asserts a different patent set and does not include '240 — but if NEC re-asserts '240 in any new filing, that filing is the event that (re)starts § 315(b) clocks for any newly named defendants and may re-open § 101 arguments that NEC has so far chosen to avoid.

Caveat on confidence: I could not locate any PTAB proceeding for US 9,953,240, and the structured ODP data affirmatively reports none. I am reporting that absence as the accurate state of the record rather than speculating about unfiled or unindexed proceedings. If you have a specific rumor of an IPR number, give it to me and I will verify it rather than guessing.

I'll research this patent's assignment record and litigation history before writing up.

Let me dig into the litigation venues and check for any post-issuance transfers.

Let me check the NEC patent-divestiture angle and confirm the absence of post-2015 assignments.

Assignment & Ownership Analysis — US 9,953,240 B2

Patent: "Image processing system, image processing method, and recording medium for detecting a static object"
App. No. 14/894,143 (371(c) national stage of PCT/JP2014/060493) · Filed 2014-04-11 · Issued 2018-04-24 · Adjusted expiry 2034-07-29 · Status Active

Sourcing note / limitation: I was able to retrieve the Google Patents legal-events record (which reproduces the USPTO assignment data) but the Assignment Center query I attempted did not return record-level detail before I exhausted my search budget. The reel/frame, conveyance dates and assignor/assignee fields below come from the Google Patents legal-events entry for this patent, not from a direct Assignment Center retrieval. The correspondent of record — the attorney/agent who filed the recording — is not surfaced in any source I could reach. I am flagging that as unknown rather than guessing, and it should be pulled directly from Assignment Center before relying on it.

Inventors

Pattern check — "inventors departing within 12 months of filing": Not observed, but the datapoint here is misleading and should not be read either way. The sole assignment was executed 2015-11-13, i.e. ~29 months after the 2013-05-31 JP priority filing and ~19 months after the 2014-04-11 PCT filing. That timing lines up exactly with national-stage entry (US app published as US20160125268A1 on 2015-05-05; the assignment was recorded 2015-11-25, the same day as the 371(c) entry event). This is the ordinary "confirm the corporate employer's title at US entry" assignment, not a departure or portfolio-liquidation tell. Note also that the JP-origin applicant (NEC) filed the PCT, so title was beneficially NEC's from the outset.

Original assignee

NEC Corporation ("NEC Corp"), 7-1, Shiba 5-chome, Minato-ku, Tokyo 108-8001, Japan.

• Primary line of business: Diversified Japanese IT/electronics — carrier networks, 5G, servers/IT services, biometrics and video-surveillance analytics, plus defense/aerospace and public-safety systems. The patent's subject matter (background-model-based detection of left-behind objects / loitering in security-camera video) sits squarely inside NEC's video-analytics and public-safety product line.
• Does it ship a product embodying the claims? Yes — this is a genuine operating-company product line, not a paper portfolio. Three independent indicators: (1) NEC pleaded in its own name that the accused eufy "package detection," AI tracking and cross-camera features read on this claim set (complaint analysis for NEC Corp. v. Anker Innovations Technology Co. Ltd., 2:24-cv-00720, ¶¶44–46, mapping Claim 1 elements to the accused functionality); (2) NEC has a long-standing admitted commercial video/biometrics business — in related litigation it "admits that it designs, develops, manufactures, and sells certain biometric identification products" (NEC's Answer in Blue Spike LLC v. NEC, E.D. Tex. 6:12-cv-690); (3) the patent's own FIG. 9 example is written around deployment in a street-installed security-camera monitoring system, i.e. an NEC-sold use case.
• Current status: Operating. No bankruptcy, receivership, IP-holding-company reorg or Chapter 7/11 on record. NEC continues to pay maintenance fees — 4th-year fee paid 2021-10-13 (M1551) and 8th-year fee paid 2025-10-16 (M1552, large entity) — and remains the assignee of record per Google Patents ("Current Assignee: NEC Corp"), with title intact through the 2034-07-29 adjusted expiry.
• Current owner of record: NEC Corporation. The chain never left the original assignee.

Assignment timeline

Exactly one recorded assignment exists. There is no post-issuance assignment — this is itself the central finding.

• 2015-11-13 (executed) / recorded 2015-11-25 — Reel 037141 / Frame 0532
  • Conveyance: Assignment
  • Assignor: EBIYAMA, Yukie (sole inventor)
  • Assignee: NEC CORPORATION (Japan)
  • Correspondent: ❗ Not determinable from the sources available to me. The correspondent of record for reel 037141/0532 was not exposed in the Google Patents legal-events feed and my Assignment Center retrieval did not complete. Because there is only one assignment link in this chain, the "repeat correspondent" test is in any event inapplicable (recurrence requires ≥2 links). Do not populate this field from any other NEC matter.
  • Context: Ordinary inventor-to-employer title confirmation, filed contemporaneously with US national-stage entry. Not an acquisition, not a transfer, not a securitization.

Related but NON-assignment record (for completeness, so it is not mis-set as a transfer):

• 2025-03-18 — Certificate of Correction ("CC" event). This is a USPTO post-grant correction of the printed patent, not a conveyance and not an ownership change. It post-dates the 2024 assertion against Anker and is consistent with clearing up patent text during that litigation.

Litigation overlay (no ownership effect — included because the family carries litigation flags):

• 2024-09-03 — NEC Corporation v. Anker Innovations Technology Co., Ltd. et al, 2:24-cv-00720 (E.D. Tex., Marshall Div.). NEC is plaintiff, asserting US 9,953,240 among six patents against Anker's "eufy" smart-home cameras, HomeBase, video doorbells and the BionicMind AI service. Reported context: NEC made an ~80-patent licensing offer to Anker in June 2024; talks failed; suit followed. NEC sought enhanced damages, fees and an injunction.
• 2024-10-24 — Power Mobile Life, LLC et al v. NEC Corporation, 1:24-cv-01870 (E.D. Va.). Anker-side plaintiffs (Power Mobile Life, LLC and Fantasia Trading, LLC d/b/a AnkerDirect) filed a declaratory-judgment action against NEC. The filing included "Exhibit A — NEC Patent Ownership," i.e. the accused infringer affirmatively litigated NEC's title/standing.
• 2024-11-01 — 1:24-cv-01870 transferred to E.D. Tex. as 2:24-cv-00881; 2024-12-05 voluntarily dismissed.
• 2024-11-03 — e-Beacon LLC v. NEC Corporation, 2:24-cv-00891 (E.D. Tex.). An NPE suing NEC as defendant; unrelated to this patent's ownership — noted only so it is not mistaken for an assertion of '240 by an NPE.

Timeline diagram

timeline
    title Ownership of US 9953240
    2013 : JP priority application filed
    2014 : PCT application filed by NEC
    2015 : Ebiyama assigns to NEC Corporation
    2016 : US application published
    2018 : Patent issued to NEC
    2024 : NEC sues Anker in East Texas
         : Anker entities file DJ action
         : DJ action transferred then dismissed
    2025 : Certificate of correction issued

NPE / troll-pattern signals

1. Shell-entity transfer — Not present. The chain contains exactly one link and it runs inventor → NEC Corporation (Reel 037141/0532, executed 2015-11-13). No "IP / Holdings / Licensing / Ventures" successor entity, no registered-agent-service address, no single-purpose LLC anywhere in the record. The patent has never left the original operating assignee.

2. Known asserter in the chain — Not present. The only assignee, NEC Corporation, appears on no NPE directory (Acacia, Marathon, IV, IPNav, Wi-LAN, Conversant/Mosaid, Vringo, Pendrell, Innovatio, MPHJ, Lumen View, Round Rock, Spangenberg entities, or Unified/RPX high-frequency-plaintiff lists). NEC is a large operating manufacturer and a frequent patent plaintiff in its own name — e.g. the June 2024 ~80-patent licensing offer to Anker followed by 2:24-cv-00720. A well-resourced operating company monetizing R&D is an operating-company assertion, categorically distinct from the NPE pattern. (Caution: NEC does divest patents — a separate NEC-origin patent, the '265 patent claiming JP 2012-048062 priority, was acquired by an entity self-described as an "intellectual property asset management" company that "acquires, manages and licenses patents" roughly five months before it sued Dell in August 2024. That is NEC-on-the-sell-side and involves a different patent; no such transfer is recorded for US 9,953,240, and I did not verify the acquirer's identity, so I treat this as context only, not as a finding about '240.)

3. Repeat correspondent across the chain — Unclear / inapplicable. There is only one recorded assignment link, so recurrence cannot be established as a matter of arithmetic. The correspondent of record for Reel 037141/0532 was not obtainable from my sources and I decline to infer it. No correspondent name should be attributed until Assignment Center is queried directly.

4. Cascading transfers — Not present. One assignment in ten-plus years of the patent's life; no consecutive chained LLC transfers, no shared correspondent addresses, no common principals. Nothing in <24 months.

5. Pre-litigation transfer — Not present. The only assignment is dated 2015-11-13, approximately 8 years and 10 months before the first suit naming this patent (2024-09-03, 2:24-cv-00720). Far outside any 6-month window; no venue-engineering or standing-cleanup transfer pattern. Notably, Anker challenged NEC's ownership via "Exhibit A — NEC Patent Ownership" in 1:24-cv-01870 — but a contested chain is not a re-arranged chain.

6. Bankruptcy fire-sale — Not present. No Chapter 7/11, no 363 sale, no distressed divestiture. NEC is an ongoing, fee-paying patentee: maintenance fees paid 2021-10-13 (4th year, M1551) and 2025-10-16 (8th year, M1552).

7. Privateering — Not present. NEC asserts in its own name as plaintiff (2:24-cv-00720), not through a proxy NPE. No SEC-disclosed transfer to a sponsored assertion vehicle, and no Patent Progress/EFF privateering coverage attaches to this patent. The chain shows no split of title that would allow a licensee-NPE to sue while NEC stays behind the curtain.

8. Defensive aggregator (anti-NPE) — Not present. No RPX, AST, LOT Network, Unified Patents or OIN assignment anywhere in the chain. The patent has not been neutralized; it remains an active assertion asset held by NEC.

Verdict

Operating-company assertion

Justification. The sole recorded assignment is inventor Yukie Ebiyama → NEC Corporation (Reel 037141/0532, executed 2015-11-13, recorded 2015-11-25), and NEC remains the assignee of record with title never having left the original operating company. NEC asserted US 9,953,240 in its own name as plaintiff against Anker's competing eufy cameras in NEC Corp. v. Anker Innovations Technology Co. Ltd., 2:24-cv-00720 (E.D. Tex., filed 2024-09-03) — after a failed ~80-patent licensing negotiation — and NEC designs, manufactures and sells video-surveillance and biometric identification products, so it is suing over a technology it actually practises. None of the eight NPE signals is present; the only adjacent noise is Anker's own declaratory-judgment counter-move (1:24-cv-01870 / 2:24-cv-00881), which contests NEC's title rather than evidencing any transfer of it.

Verify at: USPTO Assignment Center — https://assignmentcenter.uspto.gov/ (search patent number 9953240; expect a single hit at Reel 037141 / Frame 0532). Patent record: https://patents.google.com/patent/US9953240/en

Two open items for the file: (a) pull the correspondent of record for Reel 037141/0532 directly from Assignment Center — it is the one field I could not ground; (b) confirm there is no unrecorded or later-recorded transfer by checking the Assignment Center "no records found beyond" result, since an unrecorded assignment would not appear in the Google Patents legal-events feed I relied on.

I'll research US Patent 9953240 and its prior-art citations. Let me start with targeted searches.

Let me gather details on the key cited references.

Let me get details on the remaining cited references.

Prior-Art Analysis — US 9,953,240 B2

Note on sourcing: The authoritative full text of US 9,953,240 B2 and its citation lists was supplied in the task. My web searches returned the Google Patents record (which mirrors USPTO bibliographic/prosecution data) at https://patents.google.com/patent/[US9953240B2](/patent/US9953240B2)/en, confirming the bibliographic data below. I was unable to pull native USPTO PatentCenter/PAIR documents in this session, so the per-reference characterizations rely on the patent's own citation list, the reference abstracts/titles, and the full texts of the cited references that were retrievable (linked below). Where I could not verify a reference's disclosure in detail, I say so explicitly rather than speculate.

1. Target patent identification (verified, literal)

Only US 9,953,240 B2 (and its family members) is analyzed; no similar-number records were substituted.

2. Claim set that the prior art is measured against

• Claim 1 (independent, system): processor configured to (a) identify static areas from input images at multiple time points, where a static area may include motion smaller than a threshold; (b) generate first, second and third images using the static areas of images captured in first, second and third time spans from a processing time point; (c) compare the first and second images and identify an area having a difference; and (d) classify static objects according to the length of the static time period, based on comparison of the first, second and third images.
• Claim 2: second time span longer than first.
• Claim 3: generate a plurality of images for a plurality of time spans and compare them.
• Claim 4: images generated by mode/median/average of pixel values of the static areas.
• Claim 5 (independent, method): mirror of claim 1.
• Claim 6 (independent, non-transitory CRM): mirror of claim 1/5.
• Claims 7–9: method dependents paralleling claims 2–4.

Key takeaway for § 102: Independent claim 1 requires the combination of (i) isolating static (foreground-excluded) regions first, (ii) building at least three time-span images from those static regions, and (iii) classifying static objects by duration. No single cited reference that I could verify discloses all of (a)+(b)+(c)+(d) in one reference. The references below are therefore best characterized as anticipatory only for individual or clustered limitations, with the strongest § 103 relevance.

3. Cited references (13 patent citations + 1 NPL)

Reference 1 — US 2007/0280540 A1 (NEC) — highest § 102 relevance to the "static area" limitation

• Full citation: US 2007/0280540 A1, "Object detecting apparatus, method for detecting an object, and object detection program," NEC Corporation (inventor Hiroo Ikeda), pub. 2007-12-06; JP priority 2006-06-05; granted as US 8,311,273 B2 (2012-11-13). Source: https://patents.google.com/patent/US20070280540
• Description: Determines a per-pixel state from short-term variance values and a ratio R_S, classifying pixels as "stationary," "moving," "continuously stationary," or "continuously moving," and adaptively generates a background model from those states using characteristic-distance conditions; detects objects that stand still and/or move. Notably, it expressly addresses detecting an object that is not a person (e.g., something "put down and left behind").
• Potential § 102 mapping: Anticipates the "identify static areas … motion indicating a value smaller than a threshold" step of claim 1 (and the corresponding steps of claims 5 and 6), and the mean/variance background-model generation relevant to claim 4. Does not disclose generating three separate time-span images from static areas, the first-vs-second comparison per claim 1(c), or the duration-based classification of claim 1(d). Same assignee as the target patent.

Reference 2 — US 2008/0247599 A1 (Porikli) — closest overall art; this is the "PTL 1" the patent distinguishes

• Full citation: US 2008/0247599 A1, "Method for Detecting Objects Left-Behind in a Scene," Farzin M. Porikli (Mitsubishi Electric Research Laboratories), pub. 2008-10-09; priority 2007-04-05. Source: https://patents.google.com/patent/US20080247599
• Description: Analyzes pixel-intensity evolution at multiple temporal scales; maintains long-term (B_L) and short-term (B_S) background models as per-pixel mixtures of Gaussians, updated by a Bayesian mechanism; derives long-term (F_L) and short-term (F_S) foreground masks; maintains a "motion image" with evidence values/evidence rules and signals an alarm when a left-behind object is detected. Explicitly differentiates static background, active foreground, and a left-behind object that "has been static for some amount of time."
• Potential § 102 mapping: Strongest single reference for the core concept: the compare-images-from-different-time-spans limitation of claim 1(c) and the notion of discriminating objects by how long they have been static (claim 1(d) conceptually). It is the reference the specification expressly identifies as prior art to be improved upon (Background Art; the prosecution/§ 102 problem statement). It does not disclose first separating static/foreground areas and then building background images only from the static areas, nor the explicit three-time-span comparison. Best read as § 102 as to the broad comparison concept only; the target patent's asserted novelty lies in performing foreground/background separation before background-image generation.

Reference 3 — JP 5058010 B2 (Mitsubishi Electric Research Laboratories)

• Full citation: JP 5058010 B2 (granted 2012-10-24; priority 2007-04-05), "How to detect objects left behind in a scene" (ミツビシ・エレクトリック・リサーチ・ラボラトリーズ・インコーポレイテッド).
• Description: The Japanese counterpart of Reference 2 (Porikli). Same disclosure: multi-scale long-term/short-term background models compared to isolate left-behind objects.
• Potential § 102 mapping: Identical subject matter to Reference 2. This is the specific "PTL 1" cited in the target patent's Background Art; it is the primary § 102/§ 103 starting point for claim 1(c)/(d), and the patent's own arguments distinguish it on the basis of foreground/background separation and multi-window (three-span) classification.

Reference 4 — US 2013/0188826 A1 (Sony) — high relevance to time-window background generation

• Full citation: US 2013/0188826 A1, "Image processing apparatus, image processing method, and program," Sony Corporation, pub. 2013-07-25; priority 2011-08-01. Source: https://patents.google.com/patent/US20130188826
• Description: A non-moving-object detection apparatus that includes a background generation unit producing a background by weighted temporal addition over a window ΔT (a time span), a moving-object detection unit, a temporary-pause determination unit (temporarily paused images), a storage unit, and a reliability processing unit producing "non-moving object reliability C(X)"; it detects and presents non-moving objects.
• Potential § 102 mapping: Relevant to claim 1(a) (identifying static/non-moving regions) and claim 1(b) (generating a background image over a time span), and to claim 4 (temporal averaging). Does not disclose building images from only the static areas, nor the three-span classification. Its § 102(e) date (2011-08-01 priority) precedes the target's 2013-05-31 priority, so it is available as prior art.

Reference 5 — US 2017/0017845 A1 (IBM) — § 102(e) date precedes target priority

• Full citation: US 2017/0017845 A1, "Determination of object presence and motion state," International Business Machines Corp., pub. 2017-01-19; priority 2012-08-21.
• Description: Determination of object presence and motion state (i.e., distinguishing present/moving vs. stationary objects in image sequences).
• Potential § 102 mapping: Its publication date (2017) post-dates the target's grant, but its earliest effective filing (2012-08-21) precedes the target's 2013-05-31 priority, so it is potentially available under § 102(e) (pre-AIA), assuming a proper U.S. priority chain. Relevant to the "static area" determination of claim 1(a). I could not verify the detailed disclosure in this session; treat the mapping as provisional.

Reference 6 — US 2014/0056473 A1 (Canon)

• Full citation: US 2014/0056473 A1, "Object detection apparatus and control method thereof, and storage medium," Canon Kabushiki Kaisha, pub. 2014-02-27; priority 2012-08-22.
• Description: An object-detection apparatus for detecting a target object from an image; general object detection with background/foreground handling.
• Potential § 102 mapping: Its § 102(e) date (2012-08-22) precedes the target's 2013-05-31 priority. Potentially relevant to claim 1(a) (identifying a static/target region). I could not verify the detailed disclosure; the mapping is provisional and likely limited.

Reference 7 — US 2010/0231721 A1 (Searidge Technologies)

• Full citation: US 2010/0231721 A1, "Airport target tracking system," Searidge Technologies Inc., pub. 2010-09-16; priority 2007-11-30.
• Description: Airport-surface target tracking system (detection/tracking of targets in surveillance video).
• Potential § 102 mapping: Background technology in target tracking; not directed to multi-time-span static-object classification. Marginal relevance to claim 1; better as § 103 context than § 102.

Reference 8 — JP 2012-173987 A (Dainippon Printing)

• Full citation: JP 2012-173987 A, "Sunshine change area detection device and traveling object detection device," Dai Nippon Printing Co., Ltd., pub. 2012-09-10; priority 2011-02-22.
• Description: Detecting areas changed by sunlight/illumination and detecting traveling objects — i.e., handling illumination/background change.
• Potential § 102 mapping: Relevant to the problem of lighting variation that the target patent addresses (specification, "Example" section). Limited § 102 relevance to claim 1; chiefly § 103 (motivation to handle illumination changes).

Reference 9 — JP 2012-191547 A (Omron)

• Full citation: JP 2012-191547 A, "Object detection device, and object detection program," Omron Corp., pub. 2012-10-04; priority 2011-03-14.
• Description: Object-detection device/program (per title/assignee). I could not verify its detailed disclosure in this session.
• Potential § 102 mapping: Provisional; likely relevant only to claim 1(a) (object/background discrimination). Treat as § 103 context.

Reference 10 — US 2013/0128050 A1 (Aghdasi)

• Full citation: US 2013/0128050 A1, "Geographic map based control," Farzin Aghdasi, pub. 2013-05-23; priority 2011-11-22.
• Description: Video-analytics control using a geographic map (position/map-based control of analytics).
• Potential § 102 mapping: Not directed to multi-time-span static-object classification; marginal § 102 relevance to claim 1.

Reference 11 — US 2014/0064566 A1 (Xerox)

• Full citation: US 2014/0064566 A1, "Heuristic-based approach for automatic payment gesture classification and detection," Xerox Corporation, pub. 2014-03-06; priority 2012-08-29.
• Description: Heuristic gesture classification/detection (e.g., drive-thru payment gestures).
• Potential § 102 mapping: Motion/gesture classification, not static-object-by-time-span detection. Minimal § 102 relevance to claim 1; § 103 context at most.

Reference 12 — US 2015/0193661 A1 (IBM)

• Full citation: US 2015/0193661 A1, "Attribute-based alert ranking for alert adjudication," International Business Machines Corp., pub. 2015-07-09; priority 2013-01-31.
• Description: Attribute-based ranking of alerts for adjudication (surveillance alert triage).
• Potential § 102 mapping: Alert handling/ranking, not image-generation technique. Its § 102(e) date (2013-01-31) precedes the target priority (2013-05-31), so it is technically available, but its disclosure does not reach the image-generation/classification limitations of claim 1. § 103 context only.

Reference 13 — US 2015/0310365 A1 (Xerox) — date check flags a § 102 problem

• Full citation: US 2015/0310365 A1, "System and method for video-based detection of goods received event in a vehicular drive-thru," Xerox Corporation, pub. 2015-10-29; priority 2014-04-25.
• Description: Video-based detection of a goods-received event in a drive-thru.
• Potential § 102 mapping: Its stated priority date (2014-04-25) post-dates both the target's 2013-05-31 priority date and its 2014-04-11 PCT filing date. On the face of the citation list it therefore cannot be § 102(a)/(b) prior art and does not qualify under § 102(e) unless it properly claims earlier subject matter not reflected here. It appears to have been cited for a secondary/§ 103 feature rather than as an anticipatory reference. I flag this as a data-check item rather than assert anticipation.

Non-Patent Citation — International Search Report

• Full citation: International Search Report and Written Opinion dated 2014-06-24 in corresponding PCT application (PCT/JP2014/060493). Source: cited on the target patent's face.
• § 102 role: Not a substantive prior-art reference; it identifies the references the ISA considered (family members include WO 2014/192441 A1). No independent anticipation.

4. Consolidated § 102 assessment

Bottom line:

• No single cited reference anticipates independent claim 1 (or claims 5/6) in full. The two limitations that consistently defeat clean § 102 anticipation are (i) building the first/second/third images from pre-isolated static areas and (ii) classifying static objects by duration across three spans.
• US 2008/0247599 A1 / JP 5058010 B2 (Porikli/Mitsubishi) is the closest art and the reference the patent itself argues against; it is the best § 102 candidate for the multi-time-scale comparison concept underlying claim 1(c)–(d), but it lacks the static-area pre-separation and the tri-span structure.
• US 2007/0280540 A1 (NEC, Ikeda) is the best § 102 reference for the claim 1 "static area … motion smaller than a threshold" limitation and is from the same assignee (NEC).
• US 2013/0188826 A1 (Sony) is the best § 102 candidate for the "time-window background generation" limitation (claims 1(b)/4).
• Several references (US 2015/0310365 A1; the Xerox and IBM alert references) are directed to unrelated subject matter and are best treated as § 103 context; US 2015/0310365 A1 specifically post-dates the target's priority/filing and warrants a date re-check before being relied on as prior art.

Caveats: Characterizations of References 5, 6, 9, 10, and 11 rely on title/assignee/bibliographic data from the patent's citation list rather than a full-text read of each reference, which I could not complete in this session; those mappings should be verified against the references' full specifications before being relied on for an invalidity position. All § 102 dates above are as printed on the face of the target patent and are assumptions of law/date, not legal conclusions.

I'll ground the key references against live sources before building the obviousness analysis.

Let me ground the two secondary references most relevant to the multi-window/classification limitation.

Obviousness Analysis — US 9,953,240 B2 under 35 U.S.C. § 103

Analysis date: April 26, 2026
Controlling law: AIA §§ 102/103 (application filed 2014-04-11; effective filing date 2013-05-31 via JP 2013-115544)
Prior art relied on: the references appearing on the face of US 9,953,240 (Citations / Patent Citations sections), as supplemented by live verification

Two housekeeping flags before the analysis.

1. The record I have shows a Certificate of Correction dated 2025-03-18 whose contents I could not retrieve. If that correction touched claim text, every mapping below must be re-run against the corrected claims. Treat this analysis as keyed to the printed claim set.
2. The granted independent claims recite a third image and a classification step that appear neither in the printed abstract nor in the specification's Supplementary Notes (which recite only first/second images). That strongly suggests the claims were narrowed during prosecution — i.e., the examiner very likely already had Porikli-type art on the table. The obviousness case therefore has to be built specifically on the third-image and classification limitations, not on the generic background-subtraction architecture.

1. The claim, decomposed

Dependent claims add only: span ordering (2/7), "plurality of images across a plurality of time spans, compared" (3/8), and mode/median/average pixel statistics (4/9).

2. Ground 1 — Porikli alone (with NEC for the threshold limitation)

Primary reference: US 2008/0247599 A1 (Porikli & Ivanov, MERL; issued as US 7,813,528 B2), and its JP counterpart JP 5058010 B2 / EP 1 978 470 A1 — this is PTL 1, the reference the patent itself distinguishes.

Porikli discloses, expressly:

• 1.2 (foreground/background separation): "a method for detecting a left-behind object by segmenting a sequence of images … into background and foreground portions" (US 2008/0247599, ¶[0013]); foreground pixels identified by Mahalanobis distance outside a 99% confidence interval of the background Gaussians (¶[0074]). A pixel that is moving slightly but stays inside the confidence interval is treated as background — which reads directly on "motion indicating a value smaller than a threshold."
• 1.3 (multi-window background images): "The method analyzes motion in the scene at multiple temporal scales and constructs long-term and short-term background models" (¶[0016]); "we maintain parameters … of the background model at multiple time scales" (¶[0020]); "construct two background models in two parts … long-term (B_L) and short-term (B_S)" (Detailed Description). The JP 5058010 claim recites updating "a set of background models … at different temporal scales ranging from short term to long term." Critically, the MERL companion paper for the same invention (Porikli, Ivanov & Haga, Robust Abandoned Object Detection Using Dual Foregrounds, EURASIP J. Adv. Signal Process. 2008, art. 197875, §2) states: "it is possible to improve the temporal granularity by employing more than two backgrounds at different learning rates." That is an express suggestion of the third window.
• 1.4 (difference between images): the foreground masks F_L and F_S are obtained by comparing each frame against B_L and B_S, and the hypothesis table (¶ "Mask Hypothesis") distinguishes four states, including "F_L(x,y)=1 and F_S(x,y)=0 … may correspond to a left-behind object" — a difference between the two models used to locate the static object.
• 1.5 (classification by static duration): the reference's own reference point is duration: "If the B_S model 'blends' pixels into the background but the B_L model still marks the pixels as foreground, then the pixel is considered to be a left-behind object pixel. If pixel intensity changes take more time, then both models consider the pixels as background." The evidence threshold max_e "is defined in term[s] of a number of successive images, i.e., time," with a typical range [10, 300] frames — i.e., the same comparison machinery is parameterized by how long the object has been static.

Supplementary reference for the threshold-defined "static": US 2007/0280540 A1 (Ikeda, NEC; granted as US 8,311,273 B2) expressly defines pixel states including "a stationary state that is a stable state over a short period of time" and "a continuously stationary state that is a stable state over a long period of time," with object judgment based on a characteristic-distance threshold — i.e., a short-vs-long stability dichotomy and a threshold-based static determination.

Legal rationale. Where the primary reference expressly suggests more than two models at different temporal scales, adding a third window is a predictable variation of a known technique (MPEP 2144.04(IV)), not a new principle of operation. It is also an obvious design choice from a finite set of predictable solutions (KSR Int'l v. Teleflex, 550 U.S. 398, 421 (2007)). The result — a different detection boundary in the duration dimension — is the very result Porikli states it is seeking ("improve the temporal granularity").

3. Ground 2 — Porikli in view of Sony US 2013/0188826 A1 (third window + duration gating)

US 2013/0188826 A1 (Sony; EP 2 570 992 A1) discloses a non-moving-object detector in which a temporary pause determination unit "determine[s] whether the moving object is paused for a predetermined time period or more," and the "non-moving object reliability" is computed from a "temporarily paused image." Its FIG. 14 and accompanying description expressly address the case "when three temporarily paused images and three reliability images are stored in storage units 34 and 36." Sony also explicitly frames its background generation as a weighted sum over a past time window (Eq. 1).

Why combine. Both references tackle non-moving/left-behind object detection in surveillance video and both are parameterized by how long the object has been stationary. Sony supplies (a) the concrete teaching that more than two temporally-spaced stored images may be retained and used, and (b) the express duration-threshold determination that reads on claim 1.5's "classify … according to a length of a static time period." A PHOSITA seeking to distinguish, e.g., a one-minute loiterer from a ten-minute one would combine Porikli's window-generation machinery with Sony's duration-gating. Motivation is supplied by the shared problem, the shared field, and Sony's own stated objective of "robust non-moving object detection while suppressing erroneous detection."

IP status note. US 2013/0188826 published 2013-07-25 — after the 2013-05-31 critical date — so it is available only as § 102(a)(2) art; its 2011-08-01 JP priority makes it effective as such. This is the correct statutory hook and should be pleaded expressly.

4. Ground 3 — Porikli + Sony in view of IBM US 2017/0017845 A1 (classification step)

US 2017/0017845 A1 (IBM; priority 2012-08-21 → § 102(a)(2) art) labels frames by background subtraction threshold (object present/absent) and by frame-to-frame motion threshold (motion present/static), then classifies segments into "no object present," "object present and in transition," and "object present and stopped." That is a literal pedagogical disclosure of classifying a detected static object by its stationary state, using background-subtraction evidence — the closest art to claim 1.5.

Combining this with Porikli's multi-scale windows yields claim 1.5 with only the substitution of time-window index for label state as the classifier input — a mere change in the manner of using the same comparison result, and at most an obvious rearrangement of parts.

5. Ground 4 — Why these are not § 102 traps (statutory availability)

Because this is an AIA case, the availability of each reference matters and is worth stating in any petition:

6. Dependent claims (2–4, 7–9) — all obvious

• 2 / 7 (second span longer than first): Porikli is the long-term/short-term pairing. Its workable example ("short term might be a fraction of a minute, and long term five to ten minutes," train-station scenario) is a per se disclosure.
• 3 / 8 (plurality of images for a plurality of spans, compared): JP 5058010's claim of "a set of background models … updated at different temporal scales ranging from short term to long term," plus the EURASIP paper's "more than two backgrounds at different learning rates." Sony FIG. 14 (three paused images) is a redundant second showing.
• 4 / 9 (mode, median, or average): Porikli computes the expectation of the marginal posterior for the mean (Eqs. 11–13) — averaging. Sony's Eq. 1 is a weighted average. And the patent's own specification treats mode/median/average as interchangeable alternatives for a known statistical filter, which is itself a strong obviousness admission.

7. The strongest motivation argument: the patent's own admissions

The patent's specification (§ 1.3, FIGS. 5–6) concedes the failure mode of the two-window Porikli scheme and uses it as the stated reason for the invention:

"[A] target standing still for a long time, for example, longer than 10 minutes, is included in both of the background images of the long time window and the short time window, and hence detection of the target is likely to fail. … In such a case, where the long time window and the short time window have no difference, no difference is found with respect to the area actually including a static object through subsequent comparison … and hence the object is not successfully detected."

That is an express judicial admission that (i) the prior art taught every claim element except a third window, and (ii) the sole deficiency was a missing resolution along the time axis — cured by "generat[ing] background images based on multiple time windows." Under KSR and In re Kao, an admitted deficiency in the closest art plus an express suggestion in that same reference's disclosure ("more than two backgrounds") supplies both the motivation and the reasonable expectation of success. No secondary consideration is asserted anywhere in the specification (no unexpected results, no licensing, no long-felt need evidence).

8. Anticipated patent-owner rebuttals, and why they likely fail

1. "Porikli teaches away from pre-separating static areas — its whole method is pixelwise on raw frames." Weak: Porikli expressly says it works "by segmenting a sequence of images … into background and foreground portions" (¶[0013]) and computes the two foreground masks before the hypothesis logic.
2. "Sony's paused images are frame snapshots, not static-area composites; a PHOSITA could not graft it onto Porikli." This is the patent owner's best argument, and the petition should not rest on Sony for the image-generation step — Sony is offered only for the three-image/duration-gating teaching. Porikli's own "more than two backgrounds" statement, not Sony, should carry Ground 1.
3. "Classification by duration is not disclosed — Porikli only detects one left-behind object." Counter with Porikli's own text ("If pixel intensity changes take more time, then both models consider the pixels as background"; max_e measured in "number of successive images, i.e., time") and with IBM '845's express "object present and stopped" classification.
4. "Computational cost discourages more models." Porikli's cost warning concerns the number of Gaussian mixture components per pixel, not the number of time-scale backgrounds — the two must not be conflated.

9. Uncertainty and verification flags

• JP 2012-191547 A (Omron) and JP 2012-173987 A (Dainippon Printing) are listed on the face of the patent, but I could not verify their disclosures with high confidence; the last search attempt returned no usable result. JP '173987 in particular ("sunshine change area detection device") would be a useful secondary reference for the patent's lighting-variation motivation (spec § 1.7), but I do not assert its content. Do not quote these without pulling the specifications.
• Likewise, I did not verify the disclosures of US 2010/0231721 (Searidge), US 2013/0128050 (Aghdasi), US 2014/0056473 (Canon), US 2014/0064566 (Xerox), or US 2015/0193661 (IBM) beyond their bibliographic data. They are flagged as available art, not as substantively relied-upon art.
• The 2025-03-18 Certificate of Correction is unresolved. Given the abstract/specification-vs-claims mismatch I flagged in the prior section, this is the single highest-value document to retrieve before filing anything.
• § 112(b) side note: claims 5 and 6 recite "comparing the first image and the second image and identify an area having a difference" — a number/grammar mismatch. This is not an IPR ground, but it is worth noting because it may affect claim construction in the co-pending E.D. Tex. matters (2:24-cv-00720, 2:24-cv-00881) and the E.D. Va. matter (1:24-cv-01870).

10. Bottom line

Four independent § 103 grounds are available, in descending order of strength:

1. Porikli (US 2008/0247599 / JP 5058010), alone — all elements except an explicit third window, which Porikli's companion disclosure expressly suggests.
2. Porikli + NEC US 2007/0280540 — for the threshold-defined "static area" and short/long stability states.
3. Porikli + Sony US 2013/0188826 — for the third temporally-spaced image and duration-threshold gating.
4. Porikli + Sony + IBM US 2017/0017845 — for the express "classify … according to length of static time period" step.

The combination rationale is unusually strong because it rests on (a) an express suggestion in the primary reference's own family ("more than two backgrounds at different learning rates"), (b) the patent's own admitted failure of the two-window approach, and (c) no asserted objective indicia. The two genuine vulnerability points for the challenger are the retrieval of the Porikli reference's third-window teaching from the paper rather than the US patent text, and the unresolved certificate of correction.