US 9186051

I'll run searches against USPTO-style sources and CAFC docket sources for this specific patent number.

Let me run additional targeted searches on the patent number and any litigation/appeal activity.

US 9,186,051 — Analyst Summary

Scope and confidence note up front: I fetched the Google Patents record for US9186051 (and its pre-grant publication US20120114203A1) and ran targeted searches against indexed USPTO-derived, EPO/patent-family, and CAFC/litigation sources. I did not retrieve a verbatim, numbered claim set (the Google Patents page as fetched presents the claim substance inside a "Definitions" block rather than a separate claims section). The claim overview below is therefore a paraphrase of the patent's own claim/summary language, not a verbatim reproduction of claim text. Treat numeric details I could not corroborate as flagged.

1. Bibliographic data (as recorded)

Family note: A Chinese family member (CN 102469925 B) cross-lists US 13/351,342 → US 9,186,051 as granted/active, which is consistent with this being a national-phase/continuation entry off an international (PCT) chain rather than a wholly standalone filing. The US filing date being ~2.5 years after the 2009-07-23 priority date, plus the listed adjusted expiration of 2031-02-03 (implying a 20-year term running from roughly a 2011-02-03 filing), is consistent with an intervening PCT application not shown on the fetched face of the US record. I could not authoritatively confirm the specific PCT or Japanese priority application numbers from my searches, so I am not asserting them.

2. Abstract (as published)

"An image processing device includes: an interest area detector that detects interest areas included in a time-series image group captured in time series; a calculation processing unit that calculates feature amounts indicative of features of the interest areas; an area classification unit that classifies the interest areas into area groups, based on the feature amounts of the interest areas and time-series positions of time-series images including the interest areas; a group feature amount calculation unit that calculates a group feature amount indicative of a feature of each of the area groups; an area selection unit that selects one or more representative areas of the interest areas belonging to the area groups, from among the area groups; and a representative image output unit that outputs one or more representative images including the representative areas in the time-series image group."

3. Plain-language overview of the independent claims

The title and disclosure indicate three independent claims, one per statutory category (device / computer-readable recording device / method). All three share the same six-step core pipeline.

(a) Independent claim — image processing device
An image processing apparatus with six cooperating units:

1. Interest area detector — finds "interest areas" (e.g., lesions, mucosa) within a time-series image group (e.g., tens of thousands of capsule-endoscope in-vivo frames captured continuously at ~2–4 fps over ~8–10 hours).
2. Calculation processing unit (feature amount calculator) — computes feature amounts of those areas; in the described embodiment, RGB pixel values are converted to L*a*b* space and averaged to yield a color feature amount per interest area.
3. Area classification unit — groups the interest areas into "area groups" based on both (i) feature amounts and (ii) the time-series positions of the frames containing them (i.e., when each frame was captured). This is the point of novelty: similarity is judged jointly in a color feature space and along a time axis, so the same lesion seen across a run of frames collapses into one group.
4. Group feature amount calculation unit — computes a group-level feature amount describing each area group (in the embodiment, summed dispersion of the members' L*a*b* values, per Equation (2)).
5. Area selection unit — picks one or more representative areas from each area group (number of picks decided from the group feature amount via a stored function / "rate of abstract"; in one embodiment the group is sub-clustered into that many similarity groups and the area nearest each similarity-group barycenter is chosen).
6. Representative image output unit — outputs one or more representative images containing the selected representative areas from the time-series image group (e.g., to a display for a diagnosing physician).

(b) Independent claim — computer-readable recording device (program product)
The same six operations, framed as a program stored on a computer-readable recording medium that instructs a processor to perform: detecting interest areas in a time-series image group; calculating feature amounts of the interest areas; classifying the interest areas into area groups based on those feature amounts and time-series positions of the images containing them; calculating a group feature amount per area group; selecting one or more representative areas from the area groups; and outputting one or more representative images including the representative areas.

(c) Independent claim — image processing method
The same six operations recited as method steps (detecting → calculating feature amounts → classifying by feature amount and time-series position → calculating group feature amount → selecting representative area(s) → outputting representative image(s)).

Practical effect stated in the specification: interest areas that are similar in feature and close in time (e.g., the same lesion captured repeatedly within a short interval) are consolidated into a single area group, so the observer is not shown near-duplicate frames many times over — reducing reading burden and removing images containing only non-interest areas (bubbles, stool, etc.).

Likely dependent-claim subject matter (from the described embodiments, not verified as claim language): the number-of-representative-areas decision function based on group feature amount; sub-classification into similarity groups; barycenter/Euclidean-nearest selection (first embodiment); and, in the second embodiment, selection of the interest area nearest each of V equally spaced time-series coordinates T(k,i) = Tmin + W×i within a feature-point cluster.

4. Litigation / CAFC docket findings

• I found no evidence of any CAFC appeal, PTAB proceeding, or district court litigation involving US 9,186,051 in the sources I searched. The patent's record shows only prosecution, issuance, and routine assignment/address events; no security-interest, license, or litigation assignment entries.
• Important literal-matching caveat: a search for the bare string "9186051" returned an SEC EDGAR record with accession-associated identifier 9186051 for Rocky Brands, Inc. (Form 10-Q, 2013-11-06) — a numerically identical but technically unrelated identifier. This is exactly the kind of false positive that filtering on the patent number alone produces, and it is not related to this patent. Likewise, matches referencing "918" patents in the NetList (NLST) CAFC context are unrelated to US 9,186,051.
• Limits on this conclusion: absence in indexed web/patent sources is not proof of absence of litigation. Docket databases (PACER, Docket Navigator, Lex Machina) and recent 2026 CAFC filings are only partially exposed to general web search, and my final docket-specific query hit the tool's step limit before returning results. Do not treat "no litigation found" as an authoritative negative.

5. Explicit uncertainties

1. Verbatim claims not retrieved. The overview in §3 is a faithful paraphrase of the patent's own summary/definition text; exact claim numbering, transitional wording ("comprising" vs. "consisting"), and the full dependent-claim tree were not confirmed from a claims listing.
2. Priority/PCT chain unconfirmed. 2009-07-23 is listed as the priority date and Google's "prior art date," but the specific JP priority application and PCT number behind US 13/351,342 were not authoritatively identified in my searches; I am not asserting them.
3. Term calculation. The listed adjusted expiration 2031-02-03 implies a 20-year term measured from approximately 2011-02-03, which does not follow in an obvious way from a simple 2010-filed PCT national-phase entry; there may be an intermediate application. I flag this rather than resolve it.
4. Assignee identity. "Olympus Corp." (original) and "Olympus Corporation" (current) are recorded as listed; I did not verify whether any later corporate restructuring (e.g., Olympus Medical Systems Corp. entity history) affects current ownership.
5. Sibling family members beyond US 2012/0114203 A1 and CN 102469925 B (e.g., EP/JP members) were not exhaustively verified.

I'll search for litigation involving US patent 9186051.

Let me search more specifically on litigation databases.

Summary of Findings

No litigation involving U.S. Patent No. 9,186,051 has been identified.

Based on my searches of publicly available litigation and patent databases (including Google Patents/Google Patents litigation links, Unified Patents–style PTAB/litigation records surfaced via web search, Docket Alarm, CourtListener, Justia Dockets/UniCourt, and CAFC slip opinions), I found no district court case, appellate case, or PTAB post-grant proceeding (IPR/PGR/CBM) in which U.S. Patent No. 9,186,051 is the patent-in-suit or the challenged patent.

For each requested element, I therefore cannot provide a plaintiff, defendant, jurisdiction, case number, filing date, or outcome — because no such case has surfaced.

Patent identified (for confirmation that the search targeted the correct patent)

Note: The claims/abstract relate to Olympus capsule-endoscope image-processing technology, which is a field where Olympus has been a defendant in unrelated patent suits (e.g., Advanced Microscopy Inc. v. Olympus Scientific Solutions Americas, D. Del.; Display Technologies, LLC v. Olympus America, E.D. Tex. No. 2:17-cv-194; Maxell/Olympus disputes; Fleet Connect Solutions v. OM Digital Solutions/Olympus, E.D. Tex. No. 2:24-cv-01049). However, the '051 patent does not appear as an asserted patent in any of those actions — none of the results tied those cases to Patent No. 9,186,051, and I did not locate a complaint, docket entry, or judgment naming it.

Candid caveats

• A negative search result is not proof of a negative. Litigation is sometimes filed, settled, or dismissed without a searchable patent-number reference (e.g., complaints that assert a patent family only by name, sealed filings, or cases resolved pre-service). My searches returned no such indirect hit either.
• Some of my search-engine results were noise (other patents ending in "605"/"918," unrelated reexaminations, and foreign proceedings), and I expressly did not treat those as matches for US 9,186,051.
• If you need a definitive litigation clearance, the authoritative step is a direct query in Lex Machina, Darts-IP, RPX/Unified Patents Portal, or PACER (and a CAFC docket check) using the exact patent number 9,186,051; those paid/authenticated databases index every docket entry and would catch a case number I could not surface through open web search.

Bottom line: On the evidence available, there is no known litigation involving US 9,186,051.

I'll start by verifying the structured data against public sources, then probe for any proceedings the ODP ingest may have missed.

Let me probe a few more angles — litigation use, family-member patents, and patent-specific databases.

Let me run final confirmatory checks — ex parte reexam, family members, and any assertion history.

Important framing note before the count

The structured "PTAB proceedings on file" block supplied with this task states:

"The USPTO ODP API returns no AIA trial proceedings for this patent as of the most recent ingest."

My independent web research is consistent with that block. I searched for the patent number in combination with "IPR," "PGR," "PTAB," "petitioner," "reexamination certificate," and litigation terms. Every "918" hit that surfaced was a false positive for a different patent, and I want to flag those explicitly so nobody downstream repeats my search and miscites them:

I found no IPR, PGR, or CBM petition naming US 9,186,051, no institution decision, no Final Written Decision, no termination, and no Federal Circuit appeal arising from a PTAB proceeding on this patent. I was unable to complete a direct PTAB E2E / PTAB Decisions docket pull within my tool budget, so my confidence rests on the ODP block plus absence across multiple independent searches rather than a first-hand E2E query.

Proceedings overview

Total AIA trial proceedings: 0. There is no active proceeding, no proceeding in which claims were invalidated, none in which claims were sustained, none settled, and none in which institution was denied — because no petition has ever been filed. The defensive posture this gives a defendant is neither "hardened" nor "already dead": the patent is completely untested at the PTAB, which means all of its claims remain live, un-narrowed, and fully available to a plaintiff — but also means no claim has been immunized by a prior adverse or favorable Board ruling. An IPR/validity challenge here would be first-impression work with no Board precedent to build on (and, importantly, no estoppel to inherit).

Proceedings

No proceeding sections are provided, because none exist. Per the task constraints — "Do not invent proceeding numbers" and "If no PTAB activity exists, say so plainly" — I am not populating the per-proceeding template with hypotheticals.

Supporting facts about the patent itself (from the authoritative full text)

These matter for anyone deciding whether to challenge it:

• Patent: US 9,186,051 B2, "Image processing device, computer-readable recording device, and image processing method"
• Application: 13/351,342, filed 2012-01-17; granted 2015-11-17
• Priority: 2009-07-23 (JP) — i.e., pre-AIA, first-to-invent regime. This has consequences flagged below.
• Inventor: Masashi Hirota; Assignee: Olympus Corporation
• Legal status: Active; adjusted expiration 2031-02-03
• Subject matter: selecting a group of representative images from tens of thousands of time-series in-vivo images captured by a capsule endoscope — detecting interest areas, computing color feature amounts (L*a*b* averages), classifying interest areas into area groups using both feature amounts and time-series position, computing a "group feature amount" (dispersion totalized per cluster), and selecting representative areas via a rate-of-abstract function (embodiment 1: barycenter/nearest-feature-point sub-classification; embodiment 2: equal-distance time-series coordinate division).
• Claim structure (as described in the specification's summary): a device claim reciting interest area detector → calculation processing unit → area classification unit → group feature amount calculation unit → area selection unit → representative image output unit; plus a corresponding program claim and method claim. I did not retrieve the issued claim set, so I will not state claim numbers or counts — no FWD exists to quote, and I will not invent claim-level granularity.

I note one genuinely relevant drafting risk for a would-be petitioner, drawn from the spec: the classification step is expressly claimed as a two-variable operation (feature amount and time-series position), and the selection step is tied to a group feature amount (specifically totalized dispersion) feeding a rate-of-abstract. Those are narrow-ish, well-documented hooks for a § 103 obviousness attack — but also the kind of limitations that gave the applicant a clean allowance.

Strategic summary

Claim status. Every claim of US 9,186,051 is UNTESTED — none canceled, none sustained, none disclaimed, none reexamined. There is no surviving-claims analysis to perform because nothing has been stripped. If a demand letter or complaint cites any claim of this patent, that claim is fully enforceable on its face (subject to ordinary invalidity defenses in district court). Unlike the troll/patent-assertion scenario the task template anticipates, there is no "claim 1 is dead" shortcut available here.

Estoppel landscape. Because no IPR/PGR/CBM has been filed, no § 315(e)(2) estoppel attaches to anyone — not on the petitioner side and not on any privy. This cuts both ways:

• A defendant today has the complete universe of prior art available under §§ 102/103, unconstrained by any earlier petitioner's ground set or by § 325(d) "same art, same arguments" discretion (though § 325(d) would bite against art already traversed during prosecution).
• Conversely, the patent owner gains no preclusion benefit from prior bad petitioner behavior, and a defendant cannot point to an earlier IPR as "already run" to argue the PTAB has spoken.
• The one-year § 315(b) clock runs from service of a complaint alleging infringement. That is the single most important timing lever here, and it has not yet started for anyone (no litigation asserting this patent surfaced in my research).

Statutory/forum nuances specific to this patent. Because priority is 2009-07-23, the patent is pre-AIA:

• PGR is unavailable — PGR requires at least one claim with an effective filing date on/after 2013-03-16. Not possible here.
• CBM is unavailable in practice — the transitional CBM program sunset on 2020-09-16, and in any event this is a medical-imaging patent, not a "financial product or service," so it never qualified.
• IPR is the only AIA trial vehicle, and it is fully available: pre-AIA § 102/§ 103 grounds, patents and printed publications only, 35 U.S.C. §§ 311–319.
• Validity in any IPR will be judged under pre-AIA §§ 102/103, which opens doors a post-AIA petitioner would not have (e.g., pre-AIA § 102(e) art, § 102(g) prior invention/corroboration-style arguments that track the Perfect Surgical v. Olympus diligence fight).

Pattern signals. There are none on this patent: no serial petitioner, no repeated filings, no aggregator (no Unified Patents, RPX-adjacent, or similar) in the chain, no patent-owner appeal activity. Notably, Olympus is itself a frequent and successful PTAB petitioner — e.g., it was a co-petitioner in IPR2017-00714 (US 6,470,399), and it invalidated Perfect Surgical's US 6,030,384 in IPR (PTAB decision June 10, 2015). That is a signal about Olympus's sophistication, not about anyone challenging Olympus here. The more probative signal is the asymmetry: Olympus actively wields IPRs offensively, yet no one has waged one against its capsule-endoscope image-selection patent. The most likely explanations are (a) the patent has never been asserted, or (b) those who looked at it concluded a § 103 attack was not economical. Neither explanation is evidence of weakness.

Bottom line for a defendant today: you are facing a pre-AIA, actively enrolled, 2031-expiring, never-challenged patent owned by an operating company — not a troll asset with a dead claim. The absence of PTAB activity is not a defensive gift; it is the cleanest possible slate for a petitioner and the cleanest possible slate for the patent owner.

Recommended next steps

If you are a defendant and want to know what happened at the PTAB: nothing happened. There is no FWD to link to and no disposition to quote. Do not let anyone tell you otherwise. You can independently verify the null result at the USPTO PTAB E2E / PTAB Center:

• PTAB E2E / PTAB Center (search by patent number): https://ptacts.uspto.gov/ptacts/
• Patent record on Google Patents (full text + family/prosecution links): https://patents.google.com/patent/US9186051/en

Because there are no pending proceedings, there are no trial-stage milestones to track. No institution decision deadline, no oral hearing date, and no § 316(a)(11) one-year FWD date exist. If you file, the clock you care about is your own: § 315(b) gives you one year from service of an infringement complaint.

Actionable paths, in order of likely value:

1. Calendar the § 315(b) bar immediately if your client has been served. That deadline is jurisdictional in practice.
2. Commission a pre-AIA-specific prior-art search. Focus on capsule-endoscope image review/selection art published before 2009-07-23, and remember you are attacking with pre-AIA §§ 102/103 — so pre-AIA § 102(e) published applications and pre-AIA § 102(g) prior-invention theories are on the table.
3. Attack the two-step architecture as a § 103 combination. The claimed classification requires feature amount plus time-series position, and the selection requires a group feature amount feeding a rate-of-abstract. A combination of (i) an image-similarity/clustering reference, (ii) a time-series image-reduction or frame-selection reference, and (iii) a representative-frame selection reference is the natural obviousness story. The specification's own admission that it uses "publicly known clustering (refer to CG-ARTS Society, 'Digital Image Processing', p. 231) such as k-means method" is a strong motivation-to-combine hook.
4. Consider ex parte reexamination as a cheaper parallel track. It carries no § 315(e) estoppel, can be filed anonymously by a third party, and uses a preponderance standard at the CRU. Given zero PTAB history and zero § 325(d) entanglement with prior Office proceedings, a reexam request faces no "same art, same arguments" discretionary refusal.
5. Do not chase the false-positive "918" citations. If opposing counsel or a vendor report tells you this patent was invalidated, cite back the table at the top of this memo. The 918 string is heavily populated by unrelated patents, and misidentifying them is an easy and embarrassing error in a petition or an invalidity contention.

Confidence statement: I am highly confident there is no PTAB proceeding on US 9,186,051 — that conclusion is supported both by the ODP block (canonical source) and by my searches turning up nothing. I am not able to confirm, from material in front of me, whether the patent has ever been asserted in district court; that question is material to the § 315(b) analysis and should be checked against PACER / the Stanford NPE Litigation Database before relying on the "no litigation" assumption.

I'll research the ownership chain for US 9186051. Let me pull the assignment records and cross-reference.

Let me dig into the specific assignment records for this application and check for any downstream transfers.

The searches aren't surfacing reel/frame numbers directly. Let me try patent assignment aggregators and the Assignment Center index.

Ownership Analysis — US 9186051 B2

Scope note / data provenance. US 9186051 is an Olympus-origin endoscopy image-processing patent. The USPTO Assignment Center (https://assignmentcenter.uspto.gov/) was not directly reachable from this session, and my searches did not return the reel/frame numbers or correspondent-of-record fields for this patent. What follows is therefore grounded in (a) the authoritative Google Patents legal-events record for US 9186051, and (b) an INPADOC/assignment cross-check on a sibling Olympus endoscopy image-processing patent (US 9076078). All reel/frame and correspondent fields are marked [not retrieved] rather than guessed. This is an evidentiary gap, not a finding of “no records” — two assignment-type events are recorded for this patent.

Inventors

Pattern observations:

• Single-inventor, employer-assigned Japanese corporate filing — the archetypal employee-invention chain. There is no inventor-retention, no co-owner fragmentation, and no evidence of inventor departure around the transaction.
• I could not verify inventor whereabouts post-filing (no reliable source surfaced), so any “inventor departed within 12 months” thesis is unverified — do not read a fire-sale signal into it.
• No unusual patterns: the assignment was taken at filing, not later, which is inconsistent with the “inventors leave, then portfolio is sold” sequence.

Original assignee

• Entity on the issued patent: Olympus Corporation (historical name Olympus Optical Co., Ltd.; commonly styled “Olympus Corp”).
• Address of record (Olympus patents generally): 2951 Ishikawa-machi, Hachioji-shi, Tokyo 192-8507, Japan. (This address appears on Olympus EPO/validation records in Belgium’s Recueil des brevets and matches the “CHANGE OF ADDRESS” event below.)
• Primary line of business: Medical devices — flexible endoscopes, endoscopic video processors, surgical/energy devices — plus, at the relevant time, cameras/imaging, microscopes, and scientific solutions. Public company (TSE: 7733).
• Did they ship a product embodying the claims? Effectively yes. The claims cover an image-processing device that selects representative frames from a time-series in-vivo image group — functionality implemented in Olympus endoscope video-processor / image-processing software (EVIS-series class processors, lesion/mucosa detection workflows). I did not verify a specific SKU mapping this patent, so treat the product link as strong-but-not-documentary.
• Current status: Operating. Japan’s 2011 accounting-restatement episode did not result in any bankruptcy of Olympus. Note separately that Olympus’s camera/imaging division was carved out to OM Digital Solutions (majority-owned by Japan Industrial Partners) effective 2021-01-01 — but that transaction concerned the consumer/pro imaging business, not the medical/endoscopy patent estate, so it does not touch this patent.

Assignment timeline

Two recorded events appear in the Google Patents legal-events record for US 9186051. Reel/frame and correspondent fields could not be retrieved and are marked accordingly.

• Executed on or about 2012-01-17 (contemporaneous with filing) / recorded 2012-01-17 — Reel [not retrieved], Frame [not retrieved]

  • Conveyance: Assignment of assignors’ interest (“ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS)”)
  • Assignor: HIROTA, MASASHI (sole inventor)
  • Assignee: OLYMPUS CORPORATION
  • Correspondent: [not retrieved] — no recurrence can be assessed. (Because this is a standard at-filing corporate assignment, the correspondent is expected to be Olympus’s routine prosecution counsel; unverified.)
  • Context: Ordinary employee-invention assignment taken at filing — not an acquisition, fire-sale, or reorg.

• Effective 2016-04-01 / recorded 2016-06-27 — Reel [not retrieved], Frame [not retrieved] (Google Patents labels this a “CHANGE OF ADDRESS”; assignor and assignee are both OLYMPUS CORPORATION)

  • Conveyance: Change of Address (administrative; no change in beneficial ownership)
  • Assignor: OLYMPUS CORPORATION
  • Assignee: OLYMPUS CORPORATION
  • Correspondent: [not retrieved]
  • Context: Change of name/address only — a housekeeping re-record of Olympus’s address, not a transfer of rights.
  • Cross-reference (sibling patent, flagged as non-authoritative for this patent): The parallel Olympus endoscopy image-processing patent US 9076078 carries the same 2016-06-27 change-of-address event under Reel 039344 / Frame 0502, effective 2016-04-01 (INPADOC). US 9186051’s entry is almost certainly part of that same bulk Olympus filing, but I could not confirm the identical reel/frame for this patent.

• No other recorded assignment, license, security interest, merger, or release is known for US 9186051. Status per USPTO/Google Patents: Active, adjusted expiration 2031-02-03 (filing 2012-01-17; granted 2015-11-17).

Timeline diagram

timeline
    title Ownership of US 9186051
    2009 : Priority date 23 July
    2012 : US application filed by Olympus
         : Inventor assignment recorded to Olympus
         : Application published
    2015 : Patent granted
    2016 : Olympus change of address recorded
    2031 : Adjusted expiration

NPE / troll-pattern signals

1. Shell-entity transfer — not present. No transfer to any “IP / Patents / Holdings / Ventures” entity appears. The only recorded events are the original inventor→Olympus assignment and an Olympus address change. Current assignee remains Olympus Corporation (Google Patents current-assignee field).

2. Known asserter in the chain — not present. No entity from the comparator lists (Acacia, Marathon, IV, IPNav, Wi-LAN, Mosaid/Conversant, Vringo, Pendrell, Innovatio, MPHJ, Lumen View, Round Rock, DGC, Spangenberg entities, etc.) appears at any point. Assignee is unchanged since 2012.

3. Repeat correspondent across the chain — unclear. Correspondent data was not retrievable for either event. The only “chain” is a two-link, single-owner sequence, so even with the data there is no multi-assignee pattern to detect. Not scored as a finding.

4. Cascading transfers — not present. Two recorded events span ~4.4 years, and one of them (2016-06-27) is a pure change of address. No chained LLCs, no <24-month serial transfers.

5. Pre-litigation transfer — not present. No assignment within 6 months of any infringement suit; more fundamentally, no suit naming this patent was surfaced, and ownership never moved off Olympus (Google Patents shows Olympus as current assignee). The chain was not arranged for assertion.

6. Bankruptcy fire-sale — not present. Olympus Corporation is an operating public company that did not file Chapter 7/11. The famous Kodak/Nortel/Polaroid-style asset sales do not apply.

7. Privateering — not present. No transfer from Olympus to an assertion vehicle, so there is no operating-company-behind-an-NPE structure.

8. Defensive aggregator (anti-NPE) — not present. The chain does not terminate at RPX, AST, LOT, Unified Patents, or OIN; it terminates at the original operating assignee.

Verdict

Insufficient data — with the substantive observation that the ownership chain is operating-company-retained, not NPE-driven.

Justification: The only post-issuance record is the 2016-06-27 “CHANGE OF ADDRESS” Reel [not retrieved] event in which Olympus Corporation is both assignor and assignee — a housekeeping re-record (address effective 2016-04-01), not a transfer of rights. The sole genuine conveyance is the 2012-01-17 inventor→Olympus assignment (Reel [not retrieved]) taken at filing. Because I could not retrieve reel/frame numbers or correspondents from the Assignment Center, and because there is no second-owner link to evaluate, the honest classification is Insufficient data rather than any NPE or defensive-aggregator category. If forced to characterize the substance: this is a single-owner, operating-company-held patent, and none of signals 1–8 is supported by evidence.

Verification links:

• USPTO Assignment Center: https://assignmentcenter.uspto.gov/ (search by patent number 9186051)
• Google Patents legal events: https://patents.google.com/patent/US9186051/en
• Sibling-patent cross-check (INPADOC, US 9076078): https://worldwide.espacenet.com/publicationDetails/inpadoc?CC=US&NR=2009309961A1

Caveat: I did not fabricate any reel/frame or correspondent. To close the evidentiary gap, the 2012 and 2016 recordings should be pulled directly from the Assignment Center (or USPTO Patent Public Search “Assignment” tab) to capture the reel/frame and correspondent of record.

I'll search for the patent's cited references (front-page "References Cited") and related prior art.

I found the front-page "References Cited" data. Let me pull more of the list and details on each reference.

Let me recover the complete front-page reference list and claim text.

I hit the tool's step limit while retrieving the full front-page list, so the answer below distinguishes what I actually recovered from what I could not verify. I did not fabricate any citations.

Prior art / cited references for US 9,186,051 B2

A. Target identification (search-integrity note)

This is the correct record — the front-matter string is US009186051B2 and the abstract matches the six-unit pipeline in the material already generated. I filtered out numerically similar-but-unrelated identifiers (e.g., the SEC accession "9186051" noted in the earlier section) and unrelated "918/…605" families.

Important framing caveat: references listed on a patent's face are citations, not adjudicated anticipating art. A § 102 anticipation finding requires a single reference disclosing every element of a claim. Also, I could not retrieve a verbatim, numbered claim set for US 9,186,051 (consistent with the earlier section's note). Claim mapping below therefore uses claim families — the three independent claims (device / computer-readable recording device / method) and their disclosed dependent subject matter — not confirmed claim numbers. Treat the § 102 column as an analytical screen, not a legal conclusion.

B. What I retrieved

Two independent citation sets were recovered, from different documents in the same family:

1. US front page, "References Cited" (56) — from the granted US PDF (patentimages…US9186051.pdf, page 1) and corroborating snippets. ⚠️ The retrieved extraction was elided/truncated by the search index, so the U.S. list below may be incomplete (entries could exist between the ones shown). I reconstructed it faithfully but cannot certify completeness.
2. Family-level examiner citations — the "Patent Citations" list for CN 102469925 B (same family, same specification).

C. U.S. patent documents cited (per US 9,186,051 front page)

Face-page data below (number, inventor, face date, U.S. class) is as recovered. Where I give a title/description, it is either from the recovered text or flagged as my characterization.

Legend on the face: entries marked * are flagged "cited by examiner."

D. Foreign patent documents cited (front page)

E. Non-patent literature cited (front page, "Other Publications")

F. Family-level examiner citations (CN 102469925 B "Patent Citations")

These six were surfaced as the citation list attached to the Chinese family member and are the most § 102-relevant group because at least two were examiner-classified against the same disclosure:

(Note: US 2007/0025606 A1 and US 2007/0133852 A1 appear in the family list; I did not independently confirm whether they also appear on the face of the US grant, since my front-page extraction was truncated.)

G. § 102 screen — where each reference could actually land

Because the '051 independent claims are a six-element chain, no single one of the general-purpose ROI/classification references (items 1, 2, 4–10, 12, 14, and the two NPL papers) can anticipate: each supplies at most one or two links. The real § 102 pressure points are:

1. US 7,088,850 B2 (Wei et al., Edda Technology) — best § 102 candidate.

• Read against claim 1 (device) / the recording-device claim / the method claim, it plausibly discloses: detection of lesion (interest) areas ← "interest area detector"; segmentation + diagnostic information extraction over a 4-D (temporal) dataset ← "feature amounts" and the time-series half of "classify based on feature amounts and time-series positions."
• The gap to test: whether it discloses (i) a group feature amount per area group, and (ii) selecting a representative area within a group and outputting a representative image. My reading is that the "area group / group feature amount / representative area" architecture is likely absent, so in practice this is a § 103 reference unless the examiner/defendant reads those steps into its disclosure.

2. US 2005/0197567 A1 (Qian et al.) and the Olympus display cases (JP 2005-131031; JP 2006-280792).

• These target the last two links only — "selecting representative areas" and "outputting representative images." They plausibly anticipate a display/summarization-only claim, but cannot anticipate an independent claim that requires the feature+time-series classification → group feature amount architecture.

3. The classification-cluster references (US 2007/0025606 A1; US 2007/0133852 A1; US 7,496,228 B2).

• Best used as § 103 combinations with Wei et al. for the "classify interest areas into area groups" step.

4. Likely dependent-claim art (from the disclosed embodiments):

• L*a*b* color feature amounts → US 7,496,228 B2; US 2007/0195165 A1.
• Clustering / k-means into feature-point clusters → US 6,654,728 B1; Demir et al.; US 2007/0133852 A1.
• Time-series-adjacency + color-difference (ΔE) merging → US 7,088,850 B2 (temporal tracking) is the closest.
• Equally-spaced time-series coordinate selection (second embodiment, Eq. (3)) → I found no cited reference on the recovered lists that discloses this; it is likely the most defensible dependent subject matter.

Net assessment: on the recovered lists, nothing is a clean § 102 anticipation of independent claim 1; the strongest single-reference candidate is US 7,088,850 B2, and the realistic attack is a § 103 combination of Wei et al. (spatial-temporal lesion detection) + Qian et al. and/or JP 2005-131031 / JP 2006-280792 (representative-image selection and display).

H. Explicit limits on this answer

1. Reference list may be incomplete. My front-page extraction of US 9,186,051's "References Cited" was elided by the search index between several entries (e.g., between US 6,801,645 and US 7,088,850). Additional cited U.S. patents/publications may exist on the actual face.
2. No verbatim claims. § 102 mapping is by claim family and disclosed subject matter, not by confirmed claim numbers or claim language; an anticipation analysis requires the exact claim text.
3. Titles/dates. Dates and U.S. classes shown are as recovered from the face; I have flagged my own characterizations of subject matter rather than presenting inferred titles as printed text.
4. Not legal advice. These are investigative findings from an automated search, not a validity, freedom-to-operate, or infringement opinion.

I'll research the prior art cited on this patent's record and the relevant state of the art before building the § 103 analysis.

Let me pull the citation/vetting record for this patent and confirm dates on the most relevant references.

Let me get the patent's own front-page "References Cited" list.

Obviousness Analysis — U.S. Patent No. 9,186,051 B2

Scope, method, and evidentiary limits (read first)

Three candid caveats frame everything below:

1. I could not retrieve the front-page "References Cited" list or the file wrapper for US 9,186,051. My searches for the patent's own cited-art table and for the examiner's citations did not return the face-of-patent citation list, and my final retrieval attempt hit the tool's step limit. The references analyzed below are therefore (a) the art the specification itself cites, and (b) references I independently located in the same technical field, described from their own text. This is a reconstructed § 103 case, not a reproduction of the examiner's actual rejection.
2. I do not have verbatim claim text. As noted in the prior "Patent summary" section, the Google Patents record presents claim substance inside a "Definitions" block. The element-by-element mapping below is keyed to the six-step pipeline recited in the patent's Summary/Definitions language (interest area detection → feature amount calculation → classification by feature amount and time-series position → group feature amount → representative area selection → representative image output). Element labels are my labels, not claim numbers.
3. All dates below should be re-verified against the priority documents. Where a reference's prior-art qualification turns on a filing or priority date I could not confirm to the day, I say so explicitly.

1. Governing legal framework and critical dates

Because the application was filed before 16 March 2013, pre-AIA 35 U.S.C. § 103(a) governs, with pre-AIA § 102 supplying the prior-art definitions. 35 U.S.C. § 103; KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398 (2007) (rejecting rigid teaching/suggestion/motivation test; permitting "combining prior art elements according to known methods to yield predictable results," "use of a known technique to improve similar devices in the same way," and "obvious to try"); Graham v. John Deere Co., 383 U.S. 1 (1966).

Date framework, taken from the record as fetched (and flagged where inferred):

Practical consequence: a reference qualifies under pre-AIA § 102(b) if published before the critical date one year before the U.S./international filing date (roughly 2010-02-03 to 2011-01-17, depending on whether the PCT date governs), or under pre-AIA § 102(e) if it is a U.S. patent or published U.S. application filed before 2009-07-23 (or before the applicant's actual date of invention). Section 102(e) art is the more robust lever here, and it is where the strongest reference sits.

2. The prior art

Reference A — US 2009/0309961 A1 (Miyashita; Olympus Corp.), app. US 12/485,182, filed 2009-06-16, published 2009-12-17; JP priority 2008-157223 (2008-06-16)

Source: https://trade.patenthub.cn/cpc/patent-[176732](/patent/176732)-US2009309961A1-aab9f5f7472af46ffc70fcd98690a779.html

This is the closest reference on the "representative image from capsule-endoscope sequence" concept, and it is Olympus's own earlier application by a different inventive entity (Miyashita), so it satisfies "by another" under § 102(e) while being directed to the identical problem. It discloses:

• Capsule endoscope capturing 2–4 fps for ~8 hours, up to ~60,000 in-vivo images — the same factual premise as the '051.
• A feature data calculator 51 that computes feature data of each image; an area divider 52 that divides each image into mucous membrane area, lesion area, villus area, and bubble/food-residue area by clustering pixels in a feature space, explicitly including conversion to a uniform color space such as L*a*b* (also Yuv/HSV, and DCT DC coefficients).
• A classifying unit 53 that classifies each image into a type (mucous membrane image, lesion image, mucous membrane/lesion image, bubble/food residue image) using feature data and teaching data with Euclidean distance.
• An importance level determining unit 55 that assigns a higher level of importance to images "necessary for diagnosis, for example, a lesion image" and lower to bubble/food-residue images — i.e., an express interest/non-interest discriminator.
• A representative vector calculator 56 that computes "a predetermined number of representative vectors that represent a plurality of images … based on the feature data of each image and a level of importance," plus a representative vector arrangement determining unit 57 arranging them on a distribution map (SOM).
• An express admission of the general state of the art: "an image processing apparatus is known as well that selects a predetermined number of representative images from a group of sequential images and displays the representative images," citing JP 2004-321603 and an SOM-based literature approach.
• Chronological search capability (chronological image searching unit 62) keyed to a selected representative image.

Reference B — US 2010/0165088 A1 (IntroMedic Co., Ltd.), "Apparatus and Method for Displaying Capsule Endoscope Image…"

Source: https://www.freepatentsonline.com/y2010/0165088.html

Discloses, for capsule-endoscope streams: a similar-image group forming unit that "form[s] a plurality of similar-image groups with a plurality of image frames by using the endoscope image stream"; FIG. 4 is described as "a similarity between each of neighboring image frames," and FIG. 5 as "disease/bleeding analysis data obtained by a similar-image group forming unit," with FIG. 6 as "capsule-moving speed data." A representative image frame determining unit selects one frame per group, including "the image frame which is temporally or spatially positioned at the center of each similar-image group" (also highest dynamic range, highest brightness, highest edge complexity), or synthesizes a representative frame. Neighboring frames are determined from the temporal or spatial first and last image frames of each group; the number of neighboring frames may depend on "the number of the image frames included in the similar-image group."

Date flag: I could not confirm US 2010/0165088's U.S. filing date from the retrieved text. Its publication (2010) postdates the '051 priority date, so its utility as § 103 art depends on either (i) a U.S. filing date before 2009-07-23 (making it § 102(e) art), or (ii) publication before the § 102(b) critical date (~2010-02-03 to 2011-01-17). This must be verified before relying on it as a primary reference. Its teaching is nonetheless directly probative of the level of ordinary skill.

Reference C — US 8,682,142 B2 (Given Imaging Ltd.), "System and method for editing an image stream captured in-vivo"

Source: https://www.freepatentsonline.com/[8682142](/patent/8682142).html

Discloses editing an in-vivo image stream "to reduce the viewing time or generate a relatively short preview image stream intended to include images having relatively high likelihood of interest to the reviewing health professional," by "selecting a proper subset … of the captured frames for display (or equivalently selecting a proper subset of the captured frames to delete or skip)" and/or "combining, merging, averaging, or concurrently simultaneously displaying similar frames." The predetermined criteria expressly include "a rating, measure or score of the similarity (or difference) between consecutive frames, a measure or score of pathology detections (e.g., determined based on color, texture or pattern recognition analysis of pixels in the frames)," visibility/field-of-view, estimated capsule location, and combinations thereof, with frames converted to scores and combined via weighted averages; the document gives the worked example of "scanning for suspected bleeding areas and large polyps and displaying every third frame."

Date flag: US 8,682,142 issued 2014-03-25, so it is not itself § 102 art on its face; it is characterized here as evidence of a family whose earlier publications (Given Imaging's in-vivo image-editing line, priority in the mid-2000s) would be § 102(b) art. Confirm the specific earlier family member (publication number and date) before citing it.

Admitted prior art in the '051 itself

The specification states that interest areas are detected "by performing a publicly known clustering method (refer to CG-ARTS Society, 'Digital Image Processing', p. 231) such as k-means method," and that the second embodiment's sub-classification uses "publicly known clustering such as k-means method." This is an admission that the clustering machinery the claims rely on is routine, and it supplies the "known technique" prong of several KSR rationales.

3. Element-by-element mapping — independent device claim

Element (3) is the alleged point of novelty. It is disaggregated in the record: similarity-based grouping is Reference B, and time-ordered sequence handling plus interest/lesion discrimination is References A and C. Under KSR, that is a combination of known elements, not a novel architecture.

4. Combination 1 (primary): Miyashita '961 in view of IntroMedic '088

Why a PHOSITA would combine them. Both references are in the identical field (review of capsule-endoscope in-vivo image sequences), address the identical problem (an observer cannot review tens of thousands of frames), and propose the identical solution architecture (collapse redundant frames; surface the diagnostically relevant ones). KSR expressly sanctions combining references "designed to address the same problem" in the same field. Three independent rationales apply:

• Known technique to improve a similar device in the same way. Miyashita already selects representative images at the whole-image level and already segregates lesion from bubble/stool content. IntroMedic already groups temporally neighboring similar frames and picks a per-group representative. Applying B's grouping granularity and its temporal-adjacency criterion to A's lesion-discriminating pipeline is the ordinary use of a known technique on a known device ready for improvement — one of the KSR rationales verbatim.
• Predictable result. The output of the combination — fewer displayed frames, each showing a lesion-like region, with near-duplicate frames suppressed — is exactly what each reference independently seeks. There is no new interoperability problem: both operate on the same frame-ordered data, and both produce a displayed frame subset.
• Design incentive / market pressure. Miyashita's own background section concedes that representative-image selection from sequential in-vivo images is known (citing JP 2004-321603 and SOM literature). Once that concession exists, the motivation to reduce the number of representatives per group rather than globally, and to key the group boundaries to time, follows from the desire to avoid showing the same lesion repeatedly and to avoid dropping distinct lesions.

Result: the combination teaches or suggests elements (1)–(6). If the Board or a court credits B's filing date, the combination is a clean § 103 case with no gap on any element.

5. Combination 2 (alternative, and arguably stronger on the "interest area" limitation): Given Imaging (C) + IntroMedic (B), with Miyashita (A) for the feature/color-amount detail

Reference C supplies the entire editing framework and expressly recites both similarity-between-frames and pathology-detection scores as selection criteria applied to pixels/regions; Reference B supplies grouping of temporally neighboring similar frames with a per-group representative and a group-size-dependent parameter; Reference A supplies L*a*b* conversion and Euclidean-distance feature comparison and the "predetermined number of representatives" selection step. Motivation to combine is the same same-field/same-problem rationale, reinforced by C's own statement that the different editing criteria "may be applied to filter an image stream simultaneously, concurrently, one-after-another, or interleaved" — i.e., the reference itself invites combination of its criteria, which is close to an express motivation.

6. The "group feature amount" limitation and the number-of-selections function

The claim's group feature amount (summed dispersion of member L*a*b* values, Eq. (2) of the '051) and the stored function mapping group feature amount → "rate of abstract" (number selected) are the most technically specific limitations. They are nonetheless obvious:

• Dispersion/variance within a cluster is the textbook measure of cluster compactness — a mathematical step with no independent technical effect beyond the selection it feeds (Parker v. Flook, 437 U.S. 584 (1978); In re Grams, 888 F.2d 835 (Fed. Cir. 1989) (algorithmic/arithmetic steps evaluated for their contribution to the statutory result)). The '051 itself concedes k-means clustering (which minimizes within-cluster dispersion) as publicly known.
• Scaling a count to a group's size or dispersion is taught at least in spirit by Reference B ("the number of neighboring image frames … determined by … the number of the image frames included in the similar-image group") and by Reference C ("displaying every third frame") and by Reference A (a predetermined number of representative vectors weighted by importance).
• Rounding a product to the nearest whole number is arithmetic.

Dependent-claim tier (sub-classification into as many similarity groups as the number selected; barycenter computation; Euclidean-nearest interest area): each step is either the admitted k-means technique, the mathematical definition of a centroid, or nearest-neighbor vector quantization. The motivation — selecting the most "typical" member of each sub-group — is inherent in the goal of summarization and is expressly echoed by Reference B's "temporally or spatially positioned at the center of each similar-image group."

Second embodiment (time-series coordinates T(k,i) = Tmin + W·i, with W = (Tmax − Tmin)/(V+1), and selecting the interest area nearest each coordinate): this is evenly spaced temporal subsampling of a group's span. Given a group spanning [Tmin, Tmax] and a requirement to output V representatives, dividing the span into V+1 equal intervals and taking the frame nearest each division point is a routine sampling/decimation choice with a predictable benefit (representatives spread across the group's temporal extent rather than clumped). Reference B's concern with the group's temporal first and last frames and its time-centered representative supplies the motivation. The specification's own fallback for ties (oldest vs. newest) underscores the mechanical nature of the step.

7. Anticipated counterarguments and how they fare

• "The references group whole images, not interest areas." Partially true of B and C, but not of A, whose area divider 52 divides each image into a lesion area, a mucous membrane area, and a bubble/food-residue area and computes feature data for those areas with L*a*b* conversion and Euclidean distance. The gap, if any, is narrow and bridgeable by the express teachings in A and C.
• "No reference teaches combining feature similarity with temporal adjacency in one classification decision." Reference B's similar-image groups are formed on the basis of both inter-frame similarity and temporal neighborhood (first/last frames of a group are temporal parameters; FIG. 4 is "similarity between each of neighboring image frames"). The joint criterion is therefore disclosed.
• "Unexpected results / teaching away." I found no evidence of either, and no objective indicia (commercial success, licensing, long-felt need) attributable to the claims. Consistent with the earlier Litigation summary, no litigation involving US 9,186,051 was located, so there is no record of a validity challenge, injunction, or damages award that could be invoked as a secondary consideration. Absence of litigation is not evidence of non-obviousness, and it is not evidence of obviousness either — it is simply a null result.
• Realistic best patentee argument: that the specific feedback loop — computing a group-level dispersion feature and using it to set the number of representatives drawn from that same group — is not squarely disclosed by any single reference, and that the references use global or fixed counts. This is the strongest non-obviousness position, but it is a combination of known elements with predictable results argument of exactly the type KSR disfavors, and it is defeated by Reference B's group-size-dependent parameter plus Reference C's "every third frame" and Reference A's importance-weighted predetermined count.

8. Bottom line

The claims of US 9,186,051 are, on the record I could assemble, vulnerable to a § 103 rejection. The two most defensible formulations are:

1. US 2009/0309961 A1 (Miyashita, Olympus) as primary, in view of US 2010/0165088 A1 (IntroMedic) — the primary supplies interest-area detection (lesion/mucosa vs. bubble/stool), L*a*b* feature amounts, Euclidean-distance clustering, importance weighting, and per-group representative selection; the secondary supplies temporal-neighbor-based similar-frame grouping and per-group representative frame selection. The unifying motivation — reduce the reviewing physician's burden on an ~8–10 hour, tens-of-thousands-of-frame capsule study — is stated in both references. This combination is strongest on dates because US 2009/0309961's U.S. filing date (2009-06-16) precedes the '051 priority date (2009-07-23), making it § 102(e) art by roughly five weeks.
2. Given Imaging's in-vivo image-stream-editing family (US 8,682,142 and its earlier publications) in view of IntroMedic, optionally with Miyashita for the color-space feature-amount detail. This combination is stronger on the "interest area"/pathology-detection element but depends on confirming an earlier family publication date.
3. The dependent/embodiment tier (number-of-selections function, dispersion-based group feature amount, sub-classification, barycenter/Euclidean-nearest selection; and the equally spaced time-series coordinate variant) is obvious over admitted k-means prior art plus the sampling and scaling teachings of References A–C; these limitations are unlikely to rescue the independent claims.

Confidence and uncertainty. High confidence in the technical content and internal consistency of the references I actually retrieved. Medium-low confidence in the date qualifications for US 2010/0165088 and US 8,682,142's family, and I could not verify the '051's own priority/PCT chain (the earlier "Patent summary" section flagged the same unresolved 2031-02-03 term anomaly). The definitive next steps are: (i) pull the USPTO file wrapper for 13/351,342 to obtain the examiner's actual cited art and rejection history; (ii) obtain the verbatim claims; and (iii) confirm the U.S. filing and publication dates of US 2010/0165088 and of the earliest Given Imaging family publication, since both combinations turn on those dates. Any of these could shift the analysis — most likely to strengthen rather than weaken the § 103 case, given the density of same-field, same-problem art.