US 5708767

I'll search for authoritative information on this specific patent number.

Let me run additional targeted searches on litigation and the related continuation patent.

US Patent 5,708,767 — Verification Summary

Important framing note: I ran targeted searches against Google Patents/Patentsview-derived sources, USPTO PTAB/public-information documents, and Federal Circuit docket materials. I did not find any CAFC 2026 docket, appeal, or litigation referencing 5,708,767. Below I distinguish what is verified from what I could not verify.

Bibliographic data (verified against the patent's own front-matter as reproduced on Google Patents)

Note: The patent number is also cited in the European search report for EP 1 752 891 A3 (Sony, hierarchical video camera motion transition graph), where US 5,708,767 is listed as category "Y" prior art at 13 January 1998. That EP search report also lists the family member as "NONE," i.e., no non-US family members for this specific document.

Abstract (as published)

"A new technique for extracting a hierarchical decomposition of a complex video selection for browsing purposes, combines visual and temporal information to capture the important relations within a scene and between scenes in a video, thus allowing the analysis of the underlying story structure with no a priori knowledge of the content. A general model of hierarchical scene transition graph is applied to an implementation for browsing. Video shots are first identified and a collection of key frames is used to represent each video segment. These collections are then classified according to gross visual information. A platform is built on which the video is presented as directed graphs to the user, with each category of video shots represented by a node and each edge denoting a temporal relationship between categories. The analysis and processing of video is carried out directly on the compressed videos. Preliminary tests show that the narrative structure of a video selection can be effectively captured using this technique."

Plain-language overview of the independent claims

There are two independent claims: claim 1 and claim 13. (Claim 12 reads "The method of claim 1, wherein…," so it is dependent, not independent.)

Claim 1 — the broad browsing claim. A method for browsing compressed video, with two steps:

1. Build a scene transition graph hierarchically — including producing a graph layout with at least a "main scene transition graph" whose nodes represent scenes and whose edges/transition paths are illustrated for the video sequence.
2. Browse through the resulting scene transition graph.

In plain terms: the video is segmented and organized into a node-and-edge graph showing how scenes flow into one another, and the user navigates the video by working through that graph rather than a flat timeline.

Claim 13 — the decomposition/extraction claim. A method for extracting a hierarchical decomposition of a complex video selection for browsing, comprising:

1. Identifying video shots and collecting key frames from those shots, where the key frames represent each video segment;
2. Classifying the collections of key frames according to gross visual information;
3. Building a graphical representation of the video based on (a) the classification results and (b) temporal information associated with each shot — and, as an express limitation, representing each category of video shot by a node.

In plain terms: find the shots, pick representative frames, group those frames by coarse visual appearance, then draw a graph in which each group is a node, positioned/linked according to when the shots occurred.

Dependent-claim architecture (context for the independent claims)

• Claims 2–5 (from claim 1): user highlighting of a node; a window showing shots of the highlighted node; a window showing shots in temporal order.
• Claim 6–10 (from claim 1/6/7): reducing nodes to "complete scene changes" to yield a top-level scene transition graph, with the same highlighting/window features.
• Claim 11 (from claim 1): arranging identified shots in temporal order.
• Claim 12 (from claim 1): the graph-building step specifically comprising extracting DC sequences from compressed video, detecting scene changes in the DC sequences, identifying shots from those detections, and clustering the identified shots — this is the claim that ties the browsing concept to compressed-domain (MPEG/Motion JPEG DC-image) processing.
• Claims 14–19 (from claim 13): edges denote temporal relationships between categories; use of the graph to analyze underlying story structure; compression prior to identification; grouping shots into clusters by similarity measures with user interaction via hierarchical clustering; grouping of shots by proximity values; and the graph being a directed graph.

Related/continuation document (flagged as uncertain)

Google Patents lists US 5,821,945 A ("Method and apparatus for video browsing based on content and structure," priority 1995-02-03, published 1998-10-13, The Trustees of Princeton University) as citing/relating to 5,708,767. A Taiwanese counterpart document (TWI278757B) cites "US Patent No. 5,821,945" as describing the same hierarchical scene transition graph technique. I have not verified from an authoritative USPTO continuity record whether 5,821,945 is a continuation, divisional, or continuation-in-part of 08/382,877 — treat that as unconfirmed.

CAFC 2026 docket check — explicit negative finding

• I found no Federal Circuit 2026 docket entry, appeal number, or scheduled oral argument involving 5,708,767. Searches returned only unrelated 2026 CAFC/UPC material (e.g., Gibson Dunn's March 2026 Federal Circuit update; a UPC CureVac v. Moderna case management order re EP4108769).
• The only USPTO-agency hits for the literal string "5708767" were back-end search-string logs inside PTAB/public-information petition exhibits (e.g., PTAB petition document artifacts listing ("5708767" | "6282549" | …)), i.e., the number appearing in an examiner/party search history table — not a proceeding against or involving the '767 patent itself.

Uncertainty disclosures

1. I could not access a live USPTO PatentCenter/Patent Assignment record directly; assignee, inventor, and date data are drawn from the patent's own front page as reproduced on Google Patents plus a secondary patent-leaderboard listing (Bede Liu, 11 Princeton patents, listing 5,708,767 with inventors Boon‑Lock Yeo, Minerva M. Yeung, Wayne Wolf, dated 1998-01-13).
2. The claim-13 preamble text as fetched contains apparent OCR corruption ("identifying video collecting key frames") — likely "identifying video shots and collecting key frames." I am reporting the literal OCR text and flagging the probable reading.
3. "Expired – Fee Related" is Google Patents' status label with an anticipated expiration of 2015-02-03; Google expressly disclaims that this is a legal conclusion. Given the 20-year term from the 1995-02-03 filing date, expiration is consistent regardless of fee status.
4. No CAFC 2026 activity is a negative search finding, not a certification that none exists; dockets for cases filed late in 2026 could be incompletely indexed.

I'll search litigation databases and court sources for this specific patent number.

The first search returned mostly search-string artifacts and unrelated Apple/Taction cases (different patent numbers). Let me run more targeted queries against litigation-specific databases.

Litigation search results for US Patent 5,708,767

Bottom line: I found no litigation of any kind involving US Patent 5,708,767. No district court case, no Federal Circuit appeal, no PTAB proceeding, and no ITC investigation. This is a negative finding from targeted searching, not a certification.

What I searched

The one recurring "hit" — and why it is not litigation

The literal string "5708767" appears in several PTAB petition exhibits hosted at ptacts.uspto.gov, e.g. petitions 1556542, 1556543, and 1556703. In each instance it appears inside an examiner/party search-history table, as part of a large OR-string of patent numbers — for example:

S45   109   ("5708767" | "6282549" | "6304898" | ... ).PN.

These are search-query logs reproduced as evidence in other parties' proceedings. The '767 number appears because it was typed into a prior-art search string, not because the '767 patent was ever the subject of an IPR, PGR, reexam, or court case. I want to be explicit that this is a false-positive pattern that will recur if you search the raw number.

Case table

There are no rows to populate. For each requested field:

• Plaintiff(s): none known
• Defendant(s): none known
• Jurisdiction: none known
• Case number: none known
• Filing date: none known
• Outcome / status: none known

Why "no litigation" is the expected result here (context)

This is consistent with the patent's profile from the earlier section, and three structural facts explain the absence:

1. Ownership. The patent was held by Princeton University / The Trustees of Princeton University (assignment recorded 1995-06-08). University-held patents are historically asserted far less often than corporate or NPE-held patents, and this one shows no recorded reassignment to a monetization entity.
2. Term. Priority/filing 1995-02-03; anticipated expiration 2015-02-03 (Google Patents status "Expired – Fee Related"). Any assertion window closed roughly a decade ago.
3. Family. The EP search-report entry for EP 1 752 891 A3 lists the family member as "NONE," i.e., no non-US counterparts — so foreign litigation is also not a realistic source.

Caveats and uncertainty disclosures

• I could not complete an exhaustive query against USPTO PatentCenter/PTAB caselist and PACER directly within this session (one planned PTAB query returned no usable results before the tool ceiling). My negative finding rests on the searches above plus the prior section's equally negative CAFC/litigation check.
• Absence of a hit in the aggregators I reached is not proof that no suit was ever filed. Pre-2000s infringement actions involving expired university patents are sometimes poorly indexed, and a case that settled quickly with minimal docketing could escape these searches.
• I deliberately did not attribute any Apple/Taction, Omega/CalAmp, MCOM IP, or Velos Media matter to this patent. Those searches surfaced because of broad keyword overlap, but every one of them involves a different patent number; per the instruction to interpret IDs literally, none are results for 5,708,767.
• The related document US 5,821,945 (same title, same priority date, Princeton assignee) is listed on Google Patents as citing 5,708,767, and its continuity relationship to '767 remains unverified from an authoritative USPTO record. If any litigation exists in this family, it would most plausibly attach to '945 — but I found none for that number either.

I'll verify the canonical (empty) PTAB list against live sources before writing up.

Let me run additional targeted checks to confirm the negative and rule out family-level proceedings.

PTAB / AIA Trial Analysis — US 5,708,767

Cross-reference note: I am building on the earlier verification summary's bibliographic findings (filing 1995-02-03, issue 1998-01-13, 19 claims, assignee Princeton University, "Expired – Fee Related," no CAFC activity). I found no contradiction with that record. Where I make an inference that the earlier section did not establish, I flag it as an inference.

Proceedings overview

Total AIA trial proceedings on file for US 5,708,767: zero. The canonical structured list in this prompt — sourced from the USPTO Open Data Portal (ODP) AIA-trial proceeding API — returns no IPR, PGR, or CBM proceeding, and my independent web searches produced no contrary evidence (details and ruled-out false positives below).

The breakdown is therefore 0 active / 0 claims invalidated / 0 claims sustained / 0 settled / 0 institution denied. There is no proceeding template to populate, because there is no proceeding.

Bottom-line defensive posture: This is not a "hardened by surviving IPRs" patent — it is a dead patent that was never worth attacking. The patent's 20-year term from the 1995-02-03 filing ran out on 2015-02-03 (Google Patents' own anticipated-expiration entry, consistent with the 1995 filing date and with no apparent term extension), and its status label is "Expired – Fee Related," which conventionally indicates lapse for non-payment of a maintenance fee. A defendant receiving a demand letter citing 5,708,767 today should not be planning an IPR — it should be checking the calendar. Even on the most patent-owner-favorable assumptions, the infringement/recovery window closed on 2015-02-03, which is more than six years before today (2026-09-30), meaning the § 286 lookback reaches no recoverable act. Do not spend IPR money here; spend a paragraph in a response letter.

Why the list is empty (statutory eligibility analysis)

Because the count is zero, the useful practitioner output is why — i.e., which AIA vehicles were even legally available, and for how long.

Inference I want to flag explicitly (not established fact): The "Expired – Fee Related" label, combined with the 1998-01-13 issue date, points to lapse around the 11.5-year maintenance fee window — the fee due approximately 2009-07-13, with the six-month surcharge grace period ending approximately 2010-01-13. If that is right, the patent was already lapsed roughly two and a half years before the first IPR was even available (2012-09-16), which would explain the empty docket cleanly and independently of the 2015 term expiry. I did not pull the USPTO maintenance-fee payment record to confirm this, so treat the specific lapse date as unverified. Either way — 2010 lapse or 2015-02-03 statutory expiry — the proceeding count is zero, and the difference is immaterial to a 2026 defense.

False positives identified and ruled out (diligence record)

Every web hit for a "767 patent" in a PTAB or litigation context that I examined involved a different patent. Recording these so the negative finding is auditable and so no one repeats the search:

Family-level check: I also searched for AIA activity on US 5,821,945 (the Princeton document sharing the 1995-02-03 priority date that the earlier summary flagged, with continuity unconfirmed). I found none. Flagging as a negative search finding, not a certification.

Federal Circuit: Consistent with the earlier section, I found no CAFC appeal touching either 5,708,767 or a PTAB proceeding on it. There is nothing to appeal from a proceeding that never existed.

Strategic summary

Claim status of US 5,708,767: all 19 claims are UNTESTED at the PTAB — and permanently moot. No claim has been canceled by the Board, and no claim has been sustained by the Board, because no claim was ever challenged in an AIA trial. The claims' patentability is therefore judicially and administratively unadjudicated, which superficially sounds like strength but is actually the opposite of value: the claims are expired, so the only question that matters — can they be enforced — was answered by the calendar on 2015-02-03. For completeness: claim 1 (broad compressed-video browsing via a hierarchical scene transition graph) and claim 13 (the decomposition/extraction claim, with the "gross visual information" classification and node-per-category limitation) are the two independent claims; claim 12 is dependent on claim 1 despite reciting the DC-sequence/scene-change/clustering specifics. None of them has ever been construed by the Board or by a court on the record I can find.

Estoppel landscape: there is none, because estoppel requires a final written decision. 35 U.S.C. § 315(e)(2) bars a petitioner, its RPI, and its privies from raising in a civil action any ground raised or that reasonably could have been raised in an IPR that reached an FWD. With zero FWDs, § 315(e)(2) estoppel attaches to nobody. Any defendant is entirely free to assert § 102/§ 103 over any art — including the examiner-cited references US 5,231,487 (Sony Broadcast, "Storage of video signals") and US 5,329,596 (Hitachi, "Automatic clustering method"), and the 26 non-patent references (Swain & Ballard "Color Indexing," Jain & Dubes Algorithms for Clustering Data, Arman et al. ACM Multimedia '94). In practice this is academic: invalidity is not the defense, expiration is.

Pattern signals: no campaign, no aggregator, no serial petitions. There is no evidence of a repeat petitioner (there is no petitioner at all), no sign of a defensive aggregator such as Unified Patents targeting this patent, and no PTAB appeal activity by the patent owner (consistent with the earlier section's express finding of no CAFC 2026 docket). The absence of PTAB activity here is not the usual "the patent is young / just issued / not yet asserted" explanation. It is explained by the far more terminal fact of expiration, and — on my read of the "Expired – Fee Related" label — possible lapse years before IPRs became available. Note the asymmetry: the patent is heavily cited by others (Google Patents shows 189+ "cited by" entries in the snippet I retrieved and a fuller "Cited By (347)" table, including ActiveVideo Networks, Microsoft Technology Licensing, Nielsen, and Rovi/United Video), which means it functions today as prior art against other people's patents — the mirror image of an assertion asset.

One genuinely useful technical footnote for a challenger: if anyone ever did file against this patent, note that under 37 C.F.R. § 42.100(b) as amended effective 2018-11-13, the Board applies the Phillips standard to all claims. And because the patent is expired, claim amendment is unavailable in any IPR — which would strip the patent owner of the usual fallback of narrowing to a dependent claim to survive art. Neither point is worth paying for here, but both are worth knowing.

Recommended next steps

1. Do not file an IPR. There is no proceeding to file into and no economic upside. The count is zero and the patent is expired. If you are a defendant, the correct move is a short, documented response, not a petition.

2. If you received a demand letter or complaint citing US 5,708,767, run the expiration math in writing. The operative fact is that the 20-year term from the 1995-02-03 filing expired 2015-02-03, and the "Expired – Fee Related" status indicates lapse for non-payment. Layer on 35 U.S.C. § 286's six-year damages lookback: a complaint filed today (2026-09-30) reaches back only to 2020-09-30, which is more than five years after every potentially infringing act ended. The recoverable window is empty.

3. Verify the fee-lapse date if you need it for a filing. I could not confirm the maintenance-fee payment history from a live source; my estimate (12-year/11.5-year fee due ~2009-07-13, grace to ~2010-01-13) is an inference. Confirm via USPTO Patent Center → "Fees" / maintenance-fee history for US 08/382,877, and check Patent Center's "Expiration" data. Also worth pulling the Patent Assignment abstract to confirm no post-Princeton assignment to a monetization entity — the earlier section flagged that assignee data came from the patent front page and secondary sources, not a direct ownership pull.

4. If your real interest is the technology, redirect the analysis. Given the volume and identity of citing documents, US 5,708,767 is most valuable to you as § 102/§ 103 prior art — an early 1995 disclosure of hierarchical scene transition graphs, DC-sequence processing directly on MPEG/Motion JPEG compressed video, shot clustering by color-histogram intersection and luminance projection, and node/edge graph browsing of video. That is directly on point for modern video summarization, keyframe navigation, and video-embedding patents. Consider an ex parte reexamination or invalidity contention built on this reference rather than on the patent itself.

Primary sources to cite or check:

• Patent: https://patents.google.com/patent/US5708767/en (expiration entry dated 2015-02-03; status "Expired – Fee Related")
• PTAB End-to-End (PTAB E2E) case search — the authoritative place to confirm zero AIA cases: https://ptacts.uspto.gov/ptabweb/
• USPTO PTAB Decisions: https://www.uspto.gov/patents/ptab/decisions
• CourtListener (no CAFC docket found for this patent): https://www.courtlistener.com/
• USPTO Patent Center (maintenance-fee and expiration verification): https://patentcenter.uspto.gov/

Constraint honored: I have not invented a proceeding number, a judge panel, an institution decision, or an FWD for this patent. There is nothing to invent — the AIA trial docket for US 5,708,767 is empty.

I'll research the assignment/ownership record for US 5,708,767 using targeted searches against USPTO Assignment Center-derived sources and related indexes.

Let me search the assignment-record aggregators directly for this patent number.

Ownership Analysis — US 5,708,767

Method note up front. I could not open assignmentcenter.uspto.gov or assignment.uspto.gov directly with the tools available to me. The assignment data below is drawn from (a) the patent's own front-matter/legal-events record as reproduced on Google Patents, and (b) indexed third-party mirrors of USPTO assignment and litigation data. I could not retrieve the reel/frame number or the correspondent of record for the one assignment that is on file. I have marked that gap explicitly rather than reconstructing it. The most useful next step is a direct Reel/Frame or patent-number query at https://assignment.uspto.gov/patent/index.html — the record should be a 1995 reel in the 07xxx–08xxx range, but I will not state a number I have not seen.

Inventors

Employer determination: all four were Princeton-affiliated at filing. This is not speculation from the assignee name alone — the patent itself names the "Princeton video library of politics" project (Liu, Wolf, Kulkarni, Wolfe, Kobayashi, Greenstein, Fuchs, Hsu, Arman, Liang, Digital Libraries '94, pp. 215–216) as the group's context, and three of the four inventors are co-authors on the group's own cited papers.

Unusual-pattern check — NOT triggered. The scenario you flagged (all inventors departing the original assignee within 12 months of filing, preceding a fire-sale) is not present. There is no recorded departure, no re-assignment by the inventors post-issuance, and no inventor-side assignment to anyone other than Princeton. Two of the four inventors (Yeo, Yeung) did subsequently move from academia into industrial research labs — Yeo to Hewlett-Packard Laboratories and later Google, Yeung into Intel's media/security research organization — but this is ordinary academic-to-industry attrition in a university lab, occurring years after filing, and it produced no recorded patent assignment. I flag it only so it isn't mistaken for a signal.

Original assignee

The Trustees of Princeton University (Princeton, New Jersey). Assignment recorded 1995-06-08; the patent's original assignee field reads "Princeton University," and the recorded assignment names the corporate entity "PRINCETON UNIVERSITY, TRUSTEES OF."

• Primary line of business: higher education and non-profit research. Princeton is not a video-technology vendor.
• Did it ship a product embodying the claims? No. The patent's own Conclusions describe a research prototype browser — shot identification, clustering, and graph construction implemented on an SGI Indy, processing DC-sequences extracted from MPEG/Motion JPEG, with "clustering on the 16-minute Democratic Convention 1992 sequence … performed in seconds." That is a laboratory demonstration, not a commercial product. No Princeton-branded video-browsing product, and no evidence Princeton ever offered one.
• Current status: operating. Princeton is a going concern. Third-party assignee profiles show the Trustees entity still receiving grants (most recent grant dates in 2026) and still publishing applications (through 2026), so there is no dissolution, acquisition, or bankruptcy anywhere in this chain.
• Licensing posture: Princeton operates an Office of Technology Licensing and does license patents, but I found no evidence of any license, transfer, or assertion of US 5,708,767 to or by any third party.

One relevant structural fact carried over from the earlier summary: the EP search report for EP 1 752 891 A3 (Sony) lists this patent as "Y" prior art and records the family member as "NONE" — there is no non-US counterpart to trace an ownership chain through. The US record is the whole record.

Assignment timeline

Only one assignment is recorded against US 5,708,767 in the sources I could reach.

Executed date: not captured (the assignment was executed on or before the recording date; inventor-to-employer assignments of this type are normally executed contemporaneously with the 1995-02-03 filing) / recorded 1995-06-08 — Reel not retrievable from the sources reached; verify at Assignment Center

• Conveyance: Assignment of Assignors' Interest ("ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS)")
• Assignor: Bede Liu; Boon-Lock Yeo; Minerva M. Yeung; Wayne Wolf
• Assignee: Princeton University, Trustees Of
• Correspondent: not stated in the sources I retrieved. I am not supplying a name. The 2017 Princeton quantum-photonics assignment visible on the same mirror used Saul Ewing Arnstein & Lehr LLP (Wayne, PA) as correspondent — but that is a different decade, a different technology, and a different transaction, and I will not impute it to this 1995 recording.
• Context: inventor-to-university assignment — the standard, non-adversarial first link in a university research chain. Not a fire-sale, not a reorg, not a securitization, not a transfer-to-asserter.

No further records found. Specifically, I found no second assignment, no security agreement, no merger, no change-of-name, no release, and no license recorded against US 5,708,767. If that holds under a direct Assignment Center query, it is the expected and unremarkable result for a university patent: the original assignee still owns it.

Related-document caveat (carried forward, still unconfirmed): US 5,821,945 A ("Method and apparatus for video browsing based on content and structure," published 1998-10-13, The Trustees of Princeton University) shares this patent's 1995-02-03 priority date and, per FreePatentsOnline, claims an apparatus/"video browser" set (e.g., claim 7 "means for grouping video shots…", claim 13 "displayed as a hierarchical scene transition graph"). That claim set is consistent with a continuation, but I have not verified the continuity relationship from an authoritative USPTO continuity record, and I have not verified 5,821,945's own assignment chain. Treat as unconfirmed.

Timeline diagram

timeline
    title Ownership of US 5708767
    1995 : Application filed February 3
         : Inventors assign rights to Princeton
         : Assignment recorded June 8
    1998 : Patent issued January 13
    2015 : Patent term expires February 3

NPE / troll-pattern signals

1. Shell-entity transfer — Not present. The only non-inventor assignee in the entire record is The Trustees of Princeton University, a New Jersey non-profit corporation with a decades-long operating history. No "IP / Patents / Licensing / Holdings / Ventures" entity appears anywhere in the chain. No registered-agent-service address is associated with any assignee here.

2. Known asserter in the chain — Not present. No current or prior assignee matches Acacia, Marathon, Intellectual Ventures, IPNav, Wi-LAN, Mosaid/Conversant, Vringo, Pendrell, Innovatio, MPHJ, Lumen View, Round Rock, Document Generation Corp, or any Spangenberg entity. I found no Unified Patents or RPX listing identifying this patent as an asserted asset.

3. Repeat correspondent across the chain — Unclear (cannot assess). With only one recorded assignment and no correspondent name captured, there is literally nothing to test for recurrence. A single appearance would not be a finding anyway, per your own framing; this requires the Assignment Center correspondent field, which I could not retrieve. This is the single highest-value data point still outstanding.

4. Cascading transfers — Not present. There is no sequence of assignments at all, so no chained-LLC pattern within 24 months is possible.

5. Pre-litigation transfer — Not present. There is no infringement suit to measure against. The one assignment predates issuance by nearly three years and is dated four months after filing.

6. Bankruptcy fire-sale — Not present. Princeton has not filed for bankruptcy protection; no sale of this patent in any proceeding is recorded.

7. Privateering — Not present. No operating company transferred this patent to an NPE. Princeton is the origin of the chain, not a privateer, and no downstream NPE exists.

8. Defensive aggregator (anti-NPE) — Not present in the strict sense (the chain does not terminate at RPX, AST, LOT, Unified, or OIN). However, the chain terminates at a non-asserting academic institution, which yields the same practical outcome: the patent is not held by an assertion vehicle.

One red herring, flagged so it isn't misread as a signal. The literal string "5708767" does appear in publicly posted PTAB petition exhibits — specifically the search-history tables of petitions 1556542 and 1556543 at ptacts.uspto.gov. In those tables it appears inside Boolean strings alongside "5821945", "6282549", "6304898", "6385306", "6453355" (petition 1556543, query S45 returned 109 hits; petition 1556542, query L10 returned 124 hits). This is an examiner's or party's prior-art search-log artifact — the numbers were searched against, they are not the subject of the proceeding. The surrounding petition text concerns playlists, templates, and clip sequencing, and one exhibit contains an unexplained token "INTECH PROPERTIES." I have no evidence that these petitions target US 5,708,767, and I am not treating the token as an assignment fact. It is worth one footnote only because the co-searched number set may indicate a related-family cluster worth a separate look.

Verdict

Insufficient data (only the original assignment is on record).

The complete ownership chain as reconstructed is a single link: Bede Liu, Boon-Lock Yeo, Minerva M. Yeung, and Wayne Wolf assigned US 5,708,767 to Princeton University, Trustees Of, recorded 1995-06-08, executed on or before that date — a routine inventor-to-university assignment, four months after the 1995-02-03 filing and nearly three years before the 1998-01-13 issuance. No post-issuance assignment, security interest, merger, change of name, license, or release is recorded, which means the original assignee — an operating non-profit research university, not a shell, not a known asserter, and not a defensive aggregator — still holds the patent. Zero of the eight NPE signals are affirmatively present; signal 3 (repeat correspondent) is unresolvable rather than negative, because I could not retrieve the correspondent of record. The "Expired – Fee Related" status and the 2015-02-03 anticipated expiration mean the patent is past term regardless, so no assertion is possible going forward.

Verify directly at: https://assignment.uspto.gov/patent/index.html (patent-number query for 5708767) and https://assignmentcenter.uspto.gov/. The specific fields to capture and confirm against this analysis are the Reel/Frame, the execution date, and the correspondent name and address for the 1995-06-08 recording — those three fields, and confirmation that no second recording exists, would move this verdict from "insufficient data" to a firm "original assignee retains ownership, non-asserting."

I'll verify the cited references against authoritative sources. Let me run targeted searches.

I have the two examiner-cited patent references. Let me verify the Arman reference and the continuation-family document, which bear on anticipation.

Prior-Art Analysis — US Patent 5,708,767

Scope and method note

The USPTO/Google Patents front matter for US 5,708,767 A lists exactly two (2) patent citations under the heading "Patent Citations (2)," both examiner-cited (asterisked), plus 26 non-patent citations. I verified the two patent references against FreePatentsOnline, Google Patents, Unified Patents' portal, and uspto.report. (Tool ceiling reached before I could run additional confirmatory queries; noted where that matters.)

Important threshold rule applied throughout: Section 102 anticipation requires a single reference to disclose every limitation of a claim, arranged as in the claim. A reference that discloses only some elements is not anticipatory — it is at most §103 (obviousness) art. I apply that rule literally below.

Cross-reference/fl\textbf{ag to the earlier summary section: The prior section flagged the continuity relationship of US 5,821,945 as "unconfirmed." That is now resolved: uspto.report reproduces '945's Parent Case Text — "This is a continuation of application Ser. No. 08/382,877, filed Feb. 3, 1995 now U.S. Pat. No. 5,708,767." '945 (App. No. 08/856,899, filed 1997-05-15) is therefore a continuation of the '767 application, same inventors, same assignee, common specification. See §4 below.

1. The two cited patent references

Reference A — US 5,231,487 A

Brief description. The reference is directed to apparatus for storing a digital video signal, in which the signal is compressed before storage. The compression arrangement uses a spatial two-dimensional sub-band filtering arrangement that forms data sets constituting sub-bands of the 2-D spatial-frequency domain, a quantizer that applies different amounts of quantization by sub-band (with the DC-luminance sub-band quantized less than average), and an entropy encoder; the compressed signal is then stored on a magnetic storage medium (with optional error-correction/duplication of the DC sub-band). There is no teaching of shot identification, shot clustering, a scene transition graph, nodes/edges, or browsing.

Claim-by-claim §102 analysis:

Conclusion on A: Anticipates no claim. Its only relevance is as background/§103 art for the "compressed video" limitation in claim 1's preamble and the claim 16 "compressing said video selection before said identifying step" limitation — it shows compressed-domain video handling was known, but it does not meet any full claim.

Reference B — US 5,329,596 A

Brief description. The reference discloses an automatic data-clustering method for pattern recognition (expressly applicable to speech-pattern, image-pattern, or text/character-pattern recognition). In a learning (teaching) mode it builds a tree structure of classes and subclasses together with "standard patterns," by dividing sample data until each class/subclass contains data of the same category, and registering subclass/category names and standard patterns in the tree. In a recognition mode it computes a distance/similarity between the input data and the standard patterns of the lowest-level subclasses and assigns the input data to a category (with a "softened" distance condition shown in FIG. 8B). It is thus hierarchical classification by proximity — but entirely in the abstract pattern-recognition domain; no video, no shots, no scene transition graph, no browsing.

Claim-by-claim §102 analysis:

Conclusion on B: Anticipates no claim. It is potentially relevant §103 art for the clustering-by-proximity limitation in claims 12/17/18 (the "hierarchical" tree of classes/subclasses and the distance-based similarity measure are conceptually overlapping with Jain & Dubes-style hierarchical clustering used in the '767 specification).

2. The non-patent citations (26 items) — the substantively closer art

The patent's face lists 26 NPL items. The most relevant to the claimed subject matter are the following; none, standing alone, discloses a hierarchical scene transition graph with nodes representing scenes and edges/graphical representation of temporal transitions (the core of independent claims 1 and 13).

3. Consolidated §102 anticipation map

Net finding: None of the cited references — patent or non-patent — anticipates any of claims 1–19. Every independent claim (1 and 13) requires a scene transition graph / graphical representation with nodes representing scenes and edges/transitions reflecting temporal relationships, which no cited reference discloses. The references function as §103 building blocks at most. This is consistent with the references being examiner-cited and the patent having issued over them on 1998-01-13.

4. Two adjacent documents that are not §102 prior art against '767 (important to avoid mis-mapping)

(a) US 5,821,945 A — "Method and apparatus for video browsing based on content and structure," Yeo et al., The Trustees of Princeton University, App. No. 08/856,899, filed 1997-05-15, granted 1998-10-13.
uspto.report's Parent Case Text confirms: "…a continuation of application Ser. No. 08/382,877, filed Feb. 3, 1995 now U.S. Pat. No. 5,708,767."
→ Same family, same inventors, common specification, later filing/publication. It cannot be §102 or §103 prior art against '767 (it is not "by another," and it postdates '767). Note its own reference list adds US 5,521,841 and its parent 5,708,767; do not misread the Google Patents "Cited By" listing of '945 as '945 being prior art to '767.

(b) US 5,521,841 A — "Browsing contents of a given video sequence," Arman et al., granted 1996-05-28 (family: NONE per EPO annex).
→ This is the patent counterpart of the Arman et al. ACM Multimedia '94 paper. It is highly relevant art (Rframes, scene-change pre-processing, similarity grouping). However, it appears only in '945's reference list, not on '767's face. Its 1996 grant postdates '767's 1995-02-03 filing, so it would only be reachable as pre-AIA §102(e) art (US patent granted on an application by another filed before the '767 inventors' invention date) — a determination that depends on Arman's filing/invention date, which I did not verify. Treat US 5,521,841 as the most relevant non-cited art in this space, distinct from the two references actually cited on '767.

5. Caveats and uncertainty disclosures

1. Dates. US 5,231,487's priority (1991-01-11 vs 1991-01-10) and grant date (1993-07-27 vs 1993-07-26) differ between the '767 face and Unified Patents. I reported the '767-face values first and flagged the discrepancy rather than auto-correcting.
2. US 5,329,596 filing date / inventor names. I confirmed priority 1991-09-11 and publication 1994-07-12 and the "Sakou et al." inventor listing via uspto.report, but did not independently verify the application number or exact filing date before reaching the tool ceiling.
3. NPL dates. The '767 face lists the Arman ACM Multimedia '94 item without a page-level date; I report Aug. 1994 per the companion '945 listing and secondary sources.
4. §102(e) analysis of Arman's 1996 patent. Whether US 5,521,841 qualifies as §102(e) art requires Arman's application filing date relative to the '767 invention date — not verified. It is not cited on '767's face in any event.
5. Legal status. '767 is Expired – Fee Related (anticipated expiration 2015-02-03); any §102 analysis is now of historical/validity interest only.
6. Scope. This analysis maps only the references actually cited on US 5,708,767. It is not an exhaustive novelty search; additional uncited prior art (e.g., US 5,521,841, or Tonomura's video-browsing work referenced within the Arman patent) could bear on validity and would be relevant in a full §102/§103 challenge.

I'll verify the key prior-art references cited on the patent's face before constructing the § 103 analysis.

Let me ground two additional threads: the graph/video-structure art referenced within the cited NPL, and the inventors' own contemporaneous publications.

Obviousness Analysis — US 5,708,767 under 35 U.S.C. § 103

Scope note. Because the application was filed 1995-02-03, pre‑AIA § 103 governs (AIA §§ 102/103 apply only to applications filed on or after 2013-03-16). The patent is expired (anticipated expiration 2015-02-03), so this analysis matters chiefly for (a) characterizing the art of record, (b) validity analysis of same‑family document US 5,821,945, and (c) defensive/portfolio purposes. All art citations below are drawn from the "Patent Citations (2)" and "Non‑Patent Citations (26)" sections of the US 5,708,767 Google Patents record, complemented by the applicant's own admissions in the specification (which are § 103‑usable as an admission of what was known).

1. Level of ordinary skill (Graham factor 1)

The invention sits at the intersection of compressed‑domain image processing and video database/browsing interfaces. A POSITA as of February 1995 would be a person with a B.S./M.S. in EE or CS and roughly 2–3 years' experience building video indexing/browsing or MPEG/Motion‑JPEG systems, familiar with: MPEG‑1/JPEG DCT structure (the MPEG‑1 standard was ratified in 1992); DC‑image extraction from compressed streams; histogram‑based frame comparison; and classical clustering. The intellectual level is evidenced by the non‑patent citations themselves — Arman '93/'94, Zhang '93, Swain & Ballard '91 and Jain & Dubes '88 were the ordinary working literature of the field by the 1995 filing date.

2. The prior‑art inventory (as listed on this page)

Applicant admissions also of record in the '767 specification itself:

• "Scene change detection algorithms, an algorithm to detect scene transitions using the DCT coefficients of an encoded image, and algorithms to identify both abrupt and gradual scene transitions using the DC coefficients of an encoded video sequence are known in the art."
• "It is known to use a language‑based model to match the incoming video sequence… and to use a priori models… to parse the clip."
• "The directed scene transition graph is laid out using the algorithms disclosed by Laszlo Szirmay‑Kalos…" (express admission that the graph‑display step of claim 1 uses known technique).
• "Rframes may be grouped according to various criteria… the user can select a key frame, and the system then uses various criteria to search for similar key frames and present them to the user as a group" — the applicant concedes this is prior art ("It is known…").

Not prior art (flag): the Yeo & Liu "Rapid scene analysis on compressed videos" manuscript is described in the specification as "an unpublished paper submitted to IEEE" and is authored by the inventors; an unpublished manuscript is not a "printed publication" and inventors' own work is not § 102(a) art. It cannot support a § 103 rejection. (It was later published in IEEE TCSVT vol. 5, no. 6, 1995 — after the 1995‑02‑03 filing date.) Likewise, US 5,821,945 shares the 1995‑02‑03 priority date and the Princeton assignee and is a family member, not prior art (cf. pre‑AIA § 103(c) common‑ownership treatment of § 102(e)/102(f)/102(g) art).

3. Claim 1 — the primary combination

Claim 1 elements: (A) method for browsing compressed video; (B) building hierarchically a scene transition graph, incl. (B1) providing a graph layout with at least a main scene transition graph having a plurality of nodes representing scenes with the transition paths illustrated for a video sequence; (C) browsing through the resulting scene transition graph.

* Tonomura '93 and Ueda et al., "Automatic Structure Visualization for Video Editing" (INTERCHI '93) are not among the 26 NPL of record for '767 but appear in the reference list of Arman '94 (refs. [13] and [15] there) and are 1993 printed publications. I flag them as non‑of‑record art that strengthens — but is not required for — the prima facie case.

Conclusion for claim 1: Strong prima facie obviousness. Claim 1 is conspicuously broad — it does not require DC processing, does not require any particular similarity metric, does not require graph‑construction from temporal precedence, and does not require user interaction. Its only arguably distinctive feature (B1) is a cluster→node / succession→edge display, which is the express purpose of Arman '94's management techniques and the express subject matter of the admitted Szirmay‑Kalos layout.

4. Claim 13 — the decomposition claim

Elements: (i) identifying video shots and collecting key frames representing each segment; (ii) classifying the key‑frame collections according to gross visual information; (iii) building a graphical representation based on (a) the classification results and (b) temporal information associated with each shot, with each category of shot represented by a node.

Claim 12 (dependent on claim 1) is the most vulnerable of the set, because its elements are the express admissions of the Background: extracting DC sequences from compressed video (Arman '93; US 5,231,487 for DC‑domain storage), detecting scene changes between frames of the DC sequences (admitted known: "algorithms to identify both abrupt and gradual scene transitions using the DC coefficients of an encoded video sequence are known in the art"; Zhang '93 for the metrics/thresholds), identifying shots (Zhang '93), clustering the identified video shots (Jain & Dubes; Swain & Ballard; US 5,329,596). A rejection under § 102(b)/§ 103 combining Arman '93 + Zhang '93 + Swain & Ballard is close to a per se case.

5. Dependent‑claim mapping (abbreviated)

6. Motivation to combine — the articulated rationales

1. Common field and common problem. All primary references address the same problem: enabling a user to locate material in a large video database without linear scanning (Arman '94 abstract; Zhang '93 Introduction; Smoliar & Zhang '94; Gauch '94).
2. Known‑element substitution (KSR). Flat key‑frame storyboard → node‑and‑link graph; fixed grid → hierarchical clustering; pixel‑domain → DC/compressed domain. Each is a substitution of one known technique for another with predictable results.
3. Improvement of a known device ready for improvement. Arman '94's own text identifies the shortcoming of Rframe navigation; Jain & Dubes and US 5,329,596 supply the clustering machinery to collapse "hundreds of shots"; Szirmay‑Kalos supplies the display.
4. Design incentive / obvious‑to‑try for the layout algorithm. Even if selecting a particular graph‑layout algorithm were thought non‑trivial, the specification concedes the algorithm of choice was Szirmay‑Kalos's published one — a finite, known set of solutions.
5. Predictable results / reasonable expectation of success. The patent's own "Preliminary tests" report processing a 16‑minute sequence "in seconds on an SGI Indy" using only known DC‑domain computation — i.e., the claimed approach delivered the result the art predicted, not an unexpected one.

7. Where the § 103 case is weak or contestable

I want to be explicit about the soft spots rather than overstate the rejection.

• Claim 6 ("complete scene changes" → top‑level graph). No reference of record articulates the specific reduction criterion of collapsing nodes to "complete scene changes." The claim is also functionally indefinite, so the obviousness mapping rests on hierarchical clustering generally. Expect applicant to argue this is the point of novelty.
• The "scene transition graph" as a unified data structure. The combination of (i) cluster‑level nodes, (ii) edges induced by immediate temporal precedence at the shot level, and (iii) recursive lift of those edges up the hierarchy (STG properties (2)/(3) in the specification) is not shown in a single reference of record. The prima facie case therefore requires a motivation‑to‑combine argument rather than anticipation — which is precisely the kind of case KSR endorses, but it does leave room for a "teach away/improper hindsight" rebuttal.
• Evidence limitations. I have not reviewed the file wrapper (no Office Actions, no IDS transmittal, no reasons for allowance are available in what I was given). I therefore cannot state whether the 26 NPL references were considered and overcome during prosecution, whether the claims were amended, or whether the examiner relied on any of them. Absent that, the strongest available statements are the ones above; the prosecution history could materially cut either way (e.g., if an examiner allowed over Zhang '93 + Arman '94, that is some evidence of non‑obviousness, though not binding).
• Descriptions of the NPL are drawn from published abstracts and bibliographic records, cited below, not from a full‑text read of each paper. Where my mapping relies on a specific internal teaching (e.g., Arman '94's "Rframe management techniques," Jain & Dubes's treatment of user‑controlled cluster counts), it is grounded in the abstract/record, but a full‑text review is the appropriate next step before relying on this in a brief.
• No secondary‑considerations record. I found no evidence of commercial success, licensing, or industry praise with a nexus to the claims; the large forward‑citation counts on the Google Patents page (~189 listed and 347 in the "Cited By" section) reflect the fields' growth rather than a nexus‑bearing secondary consideration, and forward citations are not, standing alone, probative of non‑obviousness.

8. Bottom line

• Claim 12 is the most exposed: its elements are recited as known in the '767 Background, and Arman '93 + Zhang '93 + Swain & Ballard (optionally + US 5,329,596) render it obvious, arguably anticipated in substance.
• Claim 1 is obvious over Arman '94 (browsing/Rframes) in view of Szirmay‑Kalos (graph layout) and Jain & Dubes or US 5,329,596 (hierarchical clustering), with Arman '93 for the compressed‑video limitation.
• Claim 13 is obvious over Arman '94 (shots + key frames) in view of Swain & Ballard and Zhang '93 (classification by gross visual information) in view of Jain & Dubes/US 5,329,596 + Szirmay‑Kalos (graphical representation with a node per category and temporal links).
• Claims 2–5, 7–11, 14–19 are obvious over the same combinations plus Arman '94's navigation/management techniques and routine interface design; claims 6 and 10's "top‑level" reduction are the only elements for which the record art does not provide a clean, direct mapping.

Sources for the non‑patent references (identifiers kept literal as listed on the '767 record): ACM DL record for Arman/Depommier/Hsu/Chiu, Content‑based browsing of video sequences, ACM Multimedia '94, 97–103, https://dlnext.acm.org/doi/10.1145/[192593](/patent/192593).[192630](/patent/192630); Springer record for Zhang/Kankanhalli/Smoliar, Automatic partitioning of full‑motion video, Multimedia Systems 1(1), 10–28 (1993); Google Patents US 5,329,596 (Hitachi, Automatic clustering method), https://patents.google.com/patent/[US5329596A](/patent/US5329596A); Google Patents US 5,231,487 (Sony Broadcast, Storage of video signals), https://patents.google.com/patent/US5231487; ACM DL record for Tonomura/Akutsu/Otsuji/Sadakata, VideoMAP and VideoSpaceIcon, CHI/INTERCHI '93, https://dlnext.acm.org/doi/10.1145/[169059](/patent/169059).[169117](/patent/169117) (non‑of‑record art). The full patent text used as the authoritative source is https://patents.google.com/patent/US5708767/en.