US 7006655

I'll research this patent across USPTO records and CAFC 2026 dockets.

Let me search more specifically for litigation and USPTO records.

Let me verify litigation status and check for any 2026 CAFC activity specifically.

Summary of US Patent 7,006,655

Searches run: USPTO/Google Patents full text, Federal Circuit 2026 oral-argument calendars, and general docket/litigation sources. Note the caveat up front: I found no 2026 CAFC docket, appeal, or IPR involving US 7,006,655. Repeated searches for the number returned only (a) the patent itself, (b) unrelated documents that happen to contain the 7-digit string "7006655" (e.g., a Korean application number 10-2023-7006655, a Lithuanian registry notice, a Brazilian case number), and (c) a June 2026 CAFC scheduled-cases list that does not mention this patent. I cannot confirm any Federal Circuit activity in 2026 for this patent, and I would treat any such claim as unsupported unless a specific docket number is produced.

Bibliographic data (from the authoritative patent record)

Abstract (as issued)

The invention provides methods, apparatus and systems for embedding and detecting an electronic watermark without requiring signals for geometric compensation. A particular embodiment uses a periodic pattern with repetition as the embedded pattern itself, which carries the information embedded as the watermark, and observes the change of the pattern's period upon detection to calculate the scaling factor. When detecting, the scaling factor is computed from the digital data that underwent scaling, and the watermark detection is then performed based on that factor.

Note: the patent's own "status" line and the Google Patents legal-status field both indicate the patent expired on 2023-08-28, so it is no longer in force as of the current date.

Plain-language overview of the independent claims

The patent has one clear theme: make a watermark's own repeated pattern do double duty as the ruler used to measure how much the image was resized. No separate calibration/registration signal is embedded.

• Claim 1 (embedding method). Prepare a digital watermark pattern that carries the payload, and embed it into the digital data two or more times repetitively. The pattern must contain a tuple of base block patterns that are orthogonal to each other and must repeat horizontally and/or vertically. The embedder uses a known repetition rate (i.e., the spacing/period is known). This is the core embed-side claim.

• Claim 5 (detection method — general). When detecting such a repetitively embedded orthogonal-base-pattern watermark, use autocorrelation to find the period at which correlation is maximized. From the change in that period vs. the known original period, derive the scaling factor applied to the frame. Then detect the embedded pattern using that scaling factor and the known repetition rate. This is the core detect-side claim.

• Claim 10 (embedding method, multi-frame data). Same as Claim 1 but scoped to digital data composed of plural frames, with the pattern embedded at least twice in one of the frames. The limitation here is "horizontal or vertical" rather than "and/or."

• Claim 12 (detection method, multi-frame data). Detects a watermark embedded in one frame of a multi-frame sequence: find the maximum-correlation period by autocorrelation within that single frame, and derive the scaling factor from the period change. (It does not recite the final "detect the pattern" step that Claim 5 does.)

• Claim 14 (MPEG-specific embedding method). For MPEG data divided into 8×8-pixel intra blocks and DCT-transformed, prepare an embedded pattern made of a tuple of 8×4-pixel base block patterns that are orthogonal to each other, and embed it two or more times in one frame, repeating horizontally or vertically. The 8×4 orthogonal-block detail is what enables detection without inverse DCT.

• Claim 15 (compressed-data embedding method). Embed a repeatedly-repeating orthogonal-base-pattern watermark in compressed digital data, where an autocorrelation peak of the repetitive pattern appears differently depending on the scaling factor — the property the detector exploits.

• Claim 16 (compressed-data detection method). Find the maximum-correlation period by autocorrelation, derive the scaling factor from the period change, then detect the embedded information by applying a pre-prepared shift pattern to a detected block (the "approximate pattern / shift instead of rescale" approach, which avoids rescaling the image or the pattern).

• Claim 18 (embedding system). Apparatus version of Claim 1: means for preparing the orthogonal-base-pattern, repetition-bearing watermark and means for embedding it twice or more, with horizontal or vertical repetition.

• Claim 21 (detection system). Apparatus version of the detection approach: means for autocorrelation-based period finding, means for deriving the scaling factor from the period change, and means for detecting the pattern using that scaling factor.

• Claim 24 (computer-readable storage medium — embedding). A program storage medium whose program performs the embedding functions of Claim 1 (orthogonal base block patterns, ≥2 repetitions, horizontal/vertical repetition).

• Claim 25 (computer-readable storage medium — detection). A program storage medium whose program performs the detection functions: autocorrelation period finding, scaling-factor derivation from period change, and pattern detection using the scaling factor.

Dependent-claim thread worth noting

• Claims 2 and 6 simply restate the "known repetition rate" idea that is already in Claims 1 and 5 respectively.
• Claims 3 and 7 add the "without requiring signals for geometric compensation" limitation; Claims 4 and 8 are article-of-manufacture claims tied to Claims 3 and 7.
• Claims 9, 13, 17, 23, 26 recite that detection of the scaling factor is performed on scaled or compressed data without decompression (i.e., no inverse DCT). Note some of these claim references look irregular on the printed face of the patent — e.g., Claim 23 reads "The method of claim 21" although Claim 21 is a system claim, and Claim 26 reads "The method of claim 25" although Claim 25 is a storage-medium claim. I am reporting these exactly as they appear; I am not correcting them. Such inconsistencies are a known source of claim-definiteness challenges, but I have no evidence of any actual challenge here.
• Claim 11 links back to Claim 1 or Claim 10 and adds "embedding said embedded pattern with alteration"; Claim 19 is the system equivalent; Claim 20 is a computer program product tied to the functions of Claim 18; Claim 22 is a computer program product tied to Claim 21.

Technical gist (one paragraph)

Existing watermark schemes resisted resizing by embedding a separate geometric-calibration signal alongside the payload, which degraded image quality, slowed detection, and — when placed in the frequency domain — required an inverse transform to read. This patent instead makes the payload pattern itself periodic and repetitive, then recovers the scale factor by autocorrelating the detected pattern and measuring how far its period shifted. Two independent refinements let this work directly on compressed data: (1) using an orthogonal tuple of 8×4 base block patterns so the autocorrelation can be computed on DCT coefficients, giving sub-block scale precision (1/h) with no inverse DCT; and (2) when the scale change is small, shifting pre-computed pattern blocks to match instead of rescaling the image or the pattern. Horizontal and vertical factors are handled independently, so aspect-ratio changes are covered.

Uncertainty flags

• No CAFC 2026 activity found. I could not locate any 2026 Federal Circuit docket, briefing, or oral-argument entry for US 7,006,655. Absence of results is not proof of absence, but I have no authoritative source showing an appeal.
• No litigation found. Searches surfaced no district-court or ITC case asserting or challenging this patent. The "Cited By" entries in the Google Patents record are later patents citing this one as prior art (e.g., Digimarc screen-watermarking and watermark-orientation-signal patents), not litigation.
• Expiry. The record states expired 2023-08-28 (after a 716-day PTA). I did not independently retrieve a Patent Center maintenance-fee or expiry confirmation beyond the sources above; treat the status as reported, not verified against the fee record.

I'll research litigation involving US patent 7006655 across patent litigation databases.

Initial searches returned mostly noise (Korean application numbers, unrelated '665 patents). Let me dig deeper into litigation-specific databases.

Still noisy. Let me target litigation databases directly and check if the patent was ever asserted.

Summary of Findings — US Patent 7,006,655

Important caveat first: My searches did not surface any confirmed litigation (district court, ITC, CAFC appeal, PTAB/IPR, or reexamination) specifically involving US 7,006,655. The searches I ran were heavily contaminated by false positives from unrelated identifiers, and I could not verify a single court case naming this patent. I am reporting that negative result explicitly rather than inventing cases.

The patent (for identification-accuracy)

• Patent: US 7,006,655 B2 — "Methods, apparatus and systems for an electronic watermark having resistance to scaling"
• Inventors: Masaaki Taniguchi; Shuhichi Shimizu; Kohichi Kamijoh; Ryoh Sugihara
• Assignee: International Business Machines Corporation (Armonk, NY)
• Appl. No.: 09/952,589 · Filed: Sept. 11, 2001 · Granted: Feb. 28, 2006
• Priority: JP 2000-276129 (Sept. 12, 2000); counterpart JP3872267B2
• Legal status (per Google Patents): Expired – Lifetime, adjusted expiration Aug. 28, 2023

Litigation

No litigation known / able to be confirmed.
Specifically, I found:

• No district court complaints naming US 7,006,655 (nothing in the CourtListener / recap results, Unified Patents, or docket aggregator hits).
• No PTAB proceedings (no IPR/PGR/CBM) and no CAFC appeal referencing this patent number.
• No ITC Section 337 investigation identified.

Why the searches were inconclusive (identifiers that produced false positives)

The literal string "7006655" and "7,006,655" recurred in many unrelated records, which I am flagging so the negative result is not mistaken for a positive one:

• Korean application KR 10-2023-7006655 — an unrelated patent-office action number appearing in Samsung/Wi-LAN standard-essential-patent IPR exhibits (e.g., IPR2025-00936).
• German utility model DE 7006655 U — an unrelated "home exercise device."
• French application FR 7006655 (pub. FR2037122B1) — an unrelated pneumatic handling device.
• Numerous unrelated US "…'665" / "…'700" patents in Delaware and other complaints (e.g., Nielsen/Gracenote, Realtime Data, Bayer v. Dow re an unrelated '665 patent).

None of these involve US 7,006,655.

Recommended verification steps (since I could not confirm)

Because this appears to be a non-litigated IBM portfolio patent (or one asserted only in confidential/licensing contexts), the authoritative checks would be:

1. USPTO PTAB — search the patent number in Patent Trial and Appeal Board proceedings.
2. PACER / CourtListener (RECAP) — full-text docket search on the patent number.
3. Unified Patents litigation portal and Docket Alarm case lists.
4. USPTO Patent Assignment / Reassignment database — to see whether IBM transferred the '655 patent to an entity that later asserted it (assignment records would precede any PAE assertion).

Bottom line: Based on the searches available, there is no litigation I can confirm involving US Patent 7,006,655. Given the contamination of the identifier space and my inability to complete all targeted database lookups, treat this as "none found" rather than a definitive "none exists," and verify through PACER/CourtListener and the PTAB directly.

I'll verify the structured block's "no proceedings" default against web sources before finalizing.

Let me run a few more targeted checks to be thorough on proceeding-number conventions and any Unified Patents activity.

One result hinted at a Unified Patents portal listing. Let me check that directly.

Proceedings overview

Total AIA trial proceedings on file for US 7,006,655: 0. The USPTO Open Data Portal (ODP) structured block returns no IPR, PGR, or CBM for this patent, and web searches surfaced no proceeding number, no Final Written Decision, and no CAFC appeal — so the breakdown across active / claims-invalidated / claims-sustained / settled / institution-denied is 0 / 0 / 0 / 0 / 0. The bottom line for a defendant: all 26 claims stand exactly as they issued on 2006-02-28, never tested at the Board — but the patent expired 2023-08-28, so the real defensive posture is not "this patent is hardened," it is "this patent is dead going forward; the only exposure is past damages inside the § 286 lookback."

No per-proceeding entries to report

There are no proceeding numbers to populate the ### {PROCEEDING_NUMBER} — {Petitioner} v. {Patent Owner} blocks with. I am not going to invent one. The searches I ran were dominated by false positives from unrelated documents carrying the literal string "7006655" or a different "'655" patent, specifically:

• IPR proceedings on other "'655" patents — e.g., a semiconductor-ECO design patent (claims 1–3, 8, 11–13, 18, 21–22 challenged on anticipation), U.S. Pat. No. 8,899,655 (IPR2023-00172), and U.S. Pat. No. 7,373,655 (IPR2019-00482, Unified Patents v. American Patents LLC, claims 5–7 held unpatentable). None of these is US 7,006,655.
• Unrelated identifier collisions — Korean application KR 10-2023-7006655 (a KIPRIS office action cited as an exhibit in Samsung IPR2025-00936), and a Lithuanian registry notice "Pranešimo Nr. 7006655."

I flag these so the negative result is not miscounted as a positive one. There is no FWD to quote, no panel to name, and no claim to characterize as canceled or sustained.

Strategic summary

Claim status. Every claim of US 7,006,655 — independent claims 1, 5, 10, 12, 14, 15, 16, 18, 21, 24, 25 and all dependents — is UNTESTED at the PTAB and uncancelled. No claim has been invalidated by the Board, so there is no surviving-claim list to build an infringement theory around and no dead-claim list to attack a demand letter with. That cuts both ways: the patent owner cannot point to a PTAB win to enhance credibility, and a defendant cannot shortcut invalidity with an FWD. Any invalidity position must be built from scratch in district court or a fresh petition.

Expiry dominates everything. The record shows an adjusted expiration of 2023-08-28 (716 days of PTA under 35 U.S.C. § 154(b)). An expired patent cannot support prospective injunctive relief, and post-expiry damages are unavailable — recovery, if any, is limited to past infringement within the six-year window of 35 U.S.C. § 286 (i.e., conduct on or after roughly 2020-09-30, running only up to the 2023-08-28 expiry). This materially changes the cost/benefit of any PTAB strategy: an IPR would cost far more than the narrowing of a past-damages case is likely worth. Confirm the fee/lapse status independently (see next steps) — I did not verify the maintenance-fee record, only the reported legal-status field.

Estoppel landscape: clean. Because no petition was ever filed, there is no § 315(e)(2) estoppel against any party, and no petitioner/privy chain to worry about. Every § 102/§ 103 ground based on patents and printed publications remains fully available, in the PTAB and in litigation. Conversely, nothing has been narrowed or disclaimed, so the full issued claim scope — including the broad "orthogonal base block patterns + ≥2 repetitions" claims 1 and 15 — is what a defendant must meet.

Pattern signals: none. There is no repeat petitioner, no defensive aggregator (no Unified Patents or similar) in the chain, and no evidence the patent owner ever pursued a PTAB appeal. A Google Patents "Cited By" entry and a Unified Patents portal page do list this patent's title alongside other IBM watermark art, but that is citation/indexing activity, not an AIA proceeding. Combined with no litigation found in the prior analysis, the picture is a non-asserted, non-challenged IBM portfolio patent that has now lapsed.

One nuance a PTAB practitioner should note on the claim set. The printed claims contain internally inconsistent references — claim 23 recites "The method of claim 21" although claim 21 is a system claim, and claim 26 recites "The method of claim 25" although claim 25 is a storage-medium claim. I report these verbatim and do not correct them. Critically for forum selection: § 112 indefiniteness is not an available IPR ground — 35 U.S.C. § 311(b) limits IPR to § 102/§ 103 on patents and printed publications. That defect is a district-court/ITC argument only. The PGR window that could have reached § 112 closed roughly nine months after issue (circa late 2006), and the CBM transitional program sunset for new petitions on 2020-09-16, so neither vehicle is available now.

Recommended next steps

• If you are facing a demand letter: the dispositive facts are expiry (2023-08-28) and a maximal § 286 recovery window of six years back from filing. Ask for the asserted claim chart and the accused acts' dates; anything after 2023-08-28 is non-actionable. There is no FWD to point to because none exists.
• Before investing in an IPR: weigh it against the past-damages ceiling. An IPR remains legally available (the patent is expired but the Board will construe under Phillips), and no § 315(b) one-year bar exists until a complaint has been served on you. But for a lapsed patent with no prospective relief, litigation-side invalidity contentions are usually the cheaper path.
• Section 112 arguments go to court, not the Board. If you want to attack the claim 23 → 21 and claim 26 → 25 reference defects, brief them as indefiniteness in district court; do not waste petition pages on them at the PTAB.
• Verify the negatives directly, since my web searches cannot be exhaustive and the identifier space is heavily contaminated:
  • PTAB E2E / Patent Trial & Appeal Board: https://ptacts.uspto.gov/ptabweb
  • PTAB decisions library: https://www.uspto.gov/patents/ptab/decisions
  • CourtListener / RECAP full-text docket search: https://www.courtlistener.com
  • Unified Patents litigation portal: https://portal.unifiedpatents.com
• Absence as a signal. Well-asserted patents eventually attract IPRs. The complete absence of PTAB activity here — combined with no litigation found in the prior analysis and a 2023 expiry — is consistent with a patent that was never enforced. Treat that as corroboration, not proof, and confirm via the sources above before relying on it.

Bottom line: There are zero AIA trial proceedings on US 7,006,655. No claim has been canceled, narrowed, or sustained at the Board; there is no estoppel and no FWD. The controlling defensive facts are the 2023-08-28 expiry and the six-year § 286 lookback, not any PTAB outcome.

I'll research the assignment chain for US 7,006,655 through USPTO records and cross-reference sources.

Assignment & Ownership Analysis — US 7,006,655 B2

Caveat up front (read before the sections below): I was able to retrieve the public legal-events record for this patent, but I could not pull the underlying USPTO Assignment Center reel/frame images or the recorded correspondent field for this specific patent within the searches available. The USPTO Assignment Center (assignmentcenter.uspto.gov) and its legacy mirror returned only unrelated IBM assignments (different application numbers) when queried. Accordingly, I am not inventing a reel/frame, an execution date, or a correspondent name. Where a field is unavailable I say so explicitly. This matches the operating rule "don't fabricate assignments."

Inventors

All four named inventors are IBM personnel; their residences are all in Japan, which places the inventive work with IBM Japan (IBM Research — Yamato / Tokyo) at the time of filing.

Unusual-pattern check: None detected. There is no evidence of inventors departing, no co-ownership, and no gap between filing (2001‑09‑11) and the inventors' assignment to IBM (recorded 2002‑01‑14). All four are listed together as assignors in the single recorded assignment — a normal corporate "assignment of assignors' interest" with no fire-sale precursor. (I cannot verify individual inventor employment contracts; the IBM-employer inference rests on the Armonk assignee, all-Japan inventor residences, and the fact that IBM Japan research outputs are routinely assigned to IBM Corp.)

Original assignee

International Business Machines Corporation, Armonk, New York (New Orchard Road, Armonk, NY 10504).

• Primary line of business: Operating technology company — enterprise hardware, software, hybrid-cloud and IT services; one of the largest holders of US patents. Publicly traded (NYSE: IBM).
• Status: Operating / active. No bankruptcy, dissolution, or acquisition. (This is a real operating company, not a shell.)
• Did IBM ship a product embodying the claims? Unclear / not evidenced. The patent is a digital-watermarking research invention (content copyright-management / scaling-robust watermarking), and IBM's Watermarking / data-hiding work was an IBM Research program rather than a standalone shipped product. I found no evidence IBM commercialized or asserted a product practicing the specific claims. I will not assert product-shipping beyond what can be documented.
• Patent status: Expired — Lifetime; adjusted expiration 2023‑08‑28 (716-day PTA). The patent is no longer in force.

Assignment timeline

The only ownership event visible in the authoritative record is the initial inventor-to-corporate assignment. No post-issuance assignment, security interest, merger, change of name, license, or release is recorded in the public legal-events chain (Google Patents legal events for US 7006655 B2).

• Executed: not retrievable from sources searched / recorded 2002‑01‑14 — Reel/Frame not retrievable from sources searched (USPTO Assignment Center images could not be pulled for this patent; the legacy assignments mirror returned only unrelated IBM filings)
  • Conveyance: Assignment ("ASSIGNMENT OF ASSIGNORS' INTEREST")
  • Assignor: KAMIJOH, KOHICHI; SHIMIZU, SHIHICHI; SUGIHARA, RYOH; TANIGUCHI, MASAAKI (all four inventors)
  • Assignee: INTERNATIONAL BUSINESS MACHINES CORPORATION, Armonk, NY
  • Correspondent: not retrievable. (Note: the prosecution attorney of record on the patent face is Scully, Scott, Murphy & Presser; Louis P. Herzberg — this is the prosecution attorney, which is a separate field from the assignment-recording correspondent and must not be conflated. IBM assignments of this era were commonly self-recorded with an "IBM Corporation" internal correspondent, but I cannot confirm that for this patent.)
  • Context: ordinary inventor-to-employer assignment executed in connection with filing (application 09/952,589 filed 2001‑09‑11; priority JP 2000‑276129, 2000‑09‑12).

No further recorded links exist. Per the task instruction: if the Assignment Center has no records beyond the original assignment, say so plainly — the searchable record shows only this single link, which normally means the original assignee (IBM) retained ownership through expiry.

Timeline diagram

timeline
    title Ownership of US 7006655
    2000 : JP priority application filed
    2001 : US application filed by IBM
    2002 : Inventors assign rights to IBM
    2006 : Patent issued to IBM
    2023 : Patent expired lifetime

NPE / troll-pattern signals

Since there is only one recorded link and it runs inventors → IBM (operating company), every shell/asserter signal is absent. I mark each below and note the evidentiary basis.

1. Shell-entity transfer — not present. The sole recorded assignee is International Business Machines Corporation, an operating public company with a real Armonk, NY corporate address. No "IP / Holdings / Ventures / Licensing" suffix LLC appears anywhere in the chain.
2. Known asserter in the chain — not present. No assignee matches Acacia, Marathon, IV, Wi‑LAN, Mosaid/Conversant, Vringo, Pendrell, Round Rock, Spangenberg entities, or any RPX/Unified high-frequency-plaintiff list. The only assignee is IBM.
3. Repeat correspondent across the chain — not present (and not assessable). With a single recorded link there is no recurrence to measure; the assignment correspondent field was not retrievable. The prosecution counsel (Scully, Scott, Murphy & Presser) is a mainstream IP firm doing both operating-company and other work — a single appearance would not be a finding even if confirmed.
4. Cascading transfers — not present. No chained LLC transfers; no multiple links at all.
5. Pre-litigation transfer — not present. There is no assignment at any point, let alone within 6 months of a suit. (Cross-reference the earlier finding: no litigation naming this patent was located.)
6. Bankruptcy fire-sale — not present. IBM never filed Chapter 7/11; no sale-in-bankruptcy record.
7. Privateering — not present. No transfer to an NPE asserting on IBM's behalf; IBM retained title.
8. Defensive aggregator (anti-NPE) — not present. Chain does not terminate at RPX, AST, LOT, Unified Patents, or OIN. Ownership simply stayed with IBM until expiry.

Verdict

Insufficient data — in the sense the rubric defines it: the record contains only the original inventor-to-assignee assignment, with no post-issuance transfers and no assertion activity. Substantively, however, the single recorded link is a clean inventor → IBM operating-company assignment (recorded 2002‑01‑14), and the patent expired 2023‑08‑28. There is no evidence of any NPE involvement, shell-entity transfer, or privateering; the absence of downstream assignments indicates IBM retained title through expiry.

Verification link: USPTO Assignment Center — search by patent number 7006655: https://assignmentcenter.uspto.gov/ (legacy mirror: https://assignment.uspto.gov/patent/index.html). These are the authoritative sources I could not fully query; the reel/frame, execution date, and correspondent fields should be confirmed there before any contrary conclusion is drawn.

Contradiction / gap flags for the prior sections

• No contradiction with the earlier "no litigation / no CADC 2026 activity" finding: the assignment record is consistent with a non-asserted, non-transferred IBM portfolio patent that quietly expired.
• Gap: the previously generated sections did not include a reel/frame either; this section could not supply one. If a later pass obtains the Assignment Center image, the two open fields to fill are the reel/frame and the assignment-recording correspondent (distinct from the prosecution counsel of record).

I'll research the prior art citations on the face of US 7,006,655. Let me search for the patent's "References Cited" section, which was not included in the patent text provided to me.

I found the "References Cited" front-page section. Let me verify the complete list and get details on each reference.

Prior-Art Analysis — US Patent 7,006,655 B2

Target patent (confirmed correct number): US 7,006,655 B2 — "Methods, apparatus and systems for an electronic watermark having resistance to scaling" — Taniguchi et al., IBM, appl. 09/952,589, filed 2001‑09‑11, granted 2006‑02‑28, priority JP 2000‑276129 (2000‑09‑12). I confirmed I am reporting on this exact patent and not a similar number.

Two housekeeping notes before the substance:

1. Date discrepancy. The task header says "Current Date: April 26, 2026," while my operating date is 2026‑09‑30. I flag it rather than silently reconcile it; nothing in the analysis below depends on it.
2. Gap in the earlier-provided text and how I filled it. The Google Patents full text supplied in the prior sections is the specification/claims only — it does not contain the front‑page "References Cited" section. I therefore retrieved the citation list from the granted‑patent PDF facsimile (http://patentimages.storage.googleapis.com/6b/ba/ea/f449e66aae723d/US7006655.pdf) and the Google Patents record (https://patents.google.com/patent/US7006655B2/en). The list below is what appears on the face of the patent. Because my only source for it is an OCR rendering of that PDF, a few date fields show obvious OCR damage (e.g., "14/2004"); I report them literally and flag them rather than auto‑correcting.

1. The complete "References Cited" list on the face of US 7,006,655

(56) U.S. Patent Documents

Several entries carry a "+" or "" marker in the facsimile. On printed US patents these superscripts tie to the "cited by examiner / cited by other" footnotes; I am reporting them as shown and not over‑reading them, since the OCR makes the exact symbol unreliable.*

Foreign Patent Documents

Other Publications

2. Reference‑by‑reference § 102 analysis

Critical framing before the mapping (please read this as part of the answer): every reference on this list was before the examiner, and the patent nevertheless issued with 26 claims. That means the examiner concluded none of them anticipated under § 102. What follows is therefore not "these references invalidate the claims," but rather "these are the references a § 102 challenge would be built on, and these are the claims each is most exposed to." Where I state a reference's content, I am relying on its title/date and the technical field it occupies (I was unable to pull and read each reference's full text within this session), so the anticipation mappings are hypotheses about where the disclosure overlaps, not verified element‑by‑element hits. I flag every place where I cannot confirm actual disclosure.

The "closest" references for the two independent cores (Claims 1 and 5)

#18 — Delaunay et al., "Generalized 2‑D Cyclic Patterns for Secret Watermark Generation" (ICIP, Sep. 2000)

• Description: Concerns constructing watermark patterns that are periodic/cyclic in two dimensions and hiding a code in them. This is the single most conceptually on‑point reference for the patent's core idea of a repeated periodic pattern as the carrier.
• Potentially anticipates: Claim 1 (embedding a repetitively‑repeating pattern carrying the payload) and Claim 15 (periodic pattern in compressed data whose autocorrelation behavior varies). A § 102 attack on Claim 1 would require Delaunay to disclose a tuple of mutually orthogonal base block patterns; on the title alone I cannot confirm it does, so this is a "relevant/possible," not a confident, anticipation. It is a strong § 103 combination candidate regardless.

#22 — Kutter, "Watermarking resisting to translation, rotation, and scaling" (SPIE, Nov. 1998)

• Description: Directly on the patent's problem — RST‑resistant watermarking. The patent's stated novelty is deriving scale from the payload pattern's own period via autocorrelation rather than a separate calibration signal.
• Potentially anticipates: Claim 5 and Claim 12 (scale‑factor detection). Whether Kutter anticipates depends entirely on whether it recovers scale by autocorrelation of a repetitive embedded pattern or by some other mechanism (e.g., a transform‑domain/synchronization template). I cannot confirm the mechanism from the citation alone; if Kutter used a separate registration/calibration signal, the patent's "without requiring signals for geometric compensation" limitation (Claims 3, 7) distinguish it.

The compressed‑domain references (relevant to Claims 9, 13, 14, 15, 16, 17, 23, 26)

#19 — Langelaar, "Low Complexity Watermarks for MPEG Compressed Video" (thesis, Feb. 2000)

• Description: Low‑complexity watermarking directly in the MPEG compressed domain — the same "no inverse DCT" territory the patent claims as an advantage.
• Potentially anticipates: the "without decompression / without inverse DCT" dependent limitations in Claims 9, 13, 17, 23, 26, and is relevant to Claim 15 (compressed‑data embedding). It does not appear to target scaling‑factor recovery, so it is unlikely to reach the independent Claims 1/5/16 core, but it is a natural § 103 secondary reference against the compressed‑domain dependents.

#20 — Kiya et al., "A Method of Inserting Binary Data into MPEG Video in the Compressed Domain" (IEICE, Aug. 1999)

• Description: Inserting binary payload data into MPEG video in the compressed domain. Same relevance profile as Langelaar for the "no decompression" limitations.
• Potentially anticipates: Claims 9, 13, 17, 23, 26 (compressed‑data, no‑decompression limitations). Not a scaling‑resistance reference.

#8 / #9 — Honsinger, US 6,665,418 B1 and US 6,678,390 B1 (both 382/100)

• Description: Kodak/Honsinger‑family data‑embedding in image data, generally involving spreading/carrier patterns and (for '418) lossless recovery of the original image. These are the references most relevant to the frequency‑domain/DCT‑domain pattern claims.
• Potentially anticipates: Claim 14 (the MPEG/8×4 orthogonal base block pattern limitation) is the obvious target if the Honsinger patterns are DCT‑domain carriers — but the specific "tuple of orthogonal 8×4 base block patterns" is a narrow recitation I cannot confirm appears in either Honsinger patent. These are better § 103 material against Claim 14 than standalone § 102 art. (Note the OCR date "14/2004" on #9 is impossible; I have not corrected it.)

#7 — Wakasu, US 6,453,053 B1 (9/2002, 382/100)

• Description: A watermark/identification‑data‑superposition patent in the same 382/100 class. No scaling‑period mechanism evident from the citation.
• Potentially anticipates: at most the generic embedding machinery of Claim 1; I find no basis to map it to the autocorrelation/scale‑recovery claims.

#10 — Jones et al., US 2002/0094082 A1 (7/2002, 380/219)

• Description: A Kodak‑family application on embedding data in images (carrier‑based). Same profile as the Honsinger references.
• Potentially anticipates: potentially Claim 14/Claim 1 overlay patterns; unverified.

The Rhoads / Brandaway / Powell "foundational watermark" references (#1–#6)

#2, #3, #6 — Rhoads, US 5,748,763 / 5,748,783 / 5,832,119 (1998)
#1, #5 — Brandaway et al., US 5,530,759 / 5,825,892 (1996/1998)
#4 — Powell et al., US 5,809,160 (9/1998)

• Description: These are the foundational spread‑spectrum / steganographic image‑watermarking patents (Digimarc‑lineage Rhoads; Brandaway's secure‑image/"digital watermark" work; Powell's in‑band auxiliary‑data encoding). They disclose embedding hidden data resilient to manipulation, but their mechanism is spread‑spectrum/carrier correlation, not periodic‑pattern period measurement.
• Potentially anticipates: essentially background art only — possibly the broadest framing of Claim 1's "preparing a pattern with information to be embedded," but not the orthogonal‑base‑block tuple, not the ≥2‑repetition requirement, and not the autocorrelation‑period scale‑recovery of Claim 5. No credible standalone § 102 hit; useful for § 103 motivation‑to‑combine.

Foreign documents (#11–#17)

3. Consolidated mapping: which claims each reference most exposes

The two references I would single out as the most relevant prior art overall are:

1. Delaunay et al. (ICIP 2000) — for the embedding‑side core (periodic/cyclic 2‑D patterns carrying a code), i.e., the Claim 1/15/18/24 family.
2. Kutter (SPIE 1998) — for the detection‑side core (resistance to scaling), i.e., the Claim 5/12/21/25 family.

Maes et al. (1999) is the reference whose very title ("Shift Invariance") maps onto the patent's Claim 16 shift‑pattern embodiment, and Langelaar and Kiya are the compressed‑domain references most relevant to the "no inverse DCT / no decompression" dependents (Claims 9, 13, 17, 23, 26).

4. Verification limits (stated so this is not mistaken for more than it is)

• Source of the citation list: OCR of the granted‑patent PDF and the Google Patents record. The date field "14/2004" for US 6,678,390 B1 is impossible and is an OCR artifact I did not correct; treat that date as unverified.
• I did not read the full text of the cited references in this session. The § 102 mappings above are grounded in each reference's title, date, and field, plus the patent's own claims — not in confirmed element‑by‑element disclosures. Any statement that a reference "potentially anticipates" a claim should be read as identifying the locus of a possible challenge, not as a substantiated anticipation finding.
• Name spellings on the face of the patent appear as "Delamay" and "Langelaat"; the standard names in the literature are Delaunay and Langelaar. I have flagged the printed spellings rather than silently correcting them, per the no‑auto‑correct rule.
• Because the patent issued with all 26 claims, the examiner's own conclusion was that none of these references anticipated; the analysis above is a prioritization of where a new § 102 challenge would focus, not a reproduction of the examiner's reasoning (which I have not retrieved from the file wrapper).
• Earlier sections recorded no litigation and no 2026 CAFC activity for this patent; nothing I found in this search changes that, and the "Cited By" entries remain later‑patent citations, not litigation.

If you want, I can next pull the file‑wrapper / PTAB record and the full text of the three references I flagged as closest (Delaunay, Kutter, Maes) to convert the "potential" § 102 mappings above into verified, element‑by‑element charts.

I'll pull the actual prior-art references cited on the face of US 7,006,655 and the surrounding art before running the §103 analysis.

Let me retrieve the actual front-page references cited on US 7,006,655 and confirm the key prior-art anchors.

Obviousness Analysis — US 7,006,655 (35 U.S.C. § 103)

Scope note first. The "Prior Art" material available on the fetched Google Patents page for US 7,006,655 consists of: (i) the prior-art keyword set ("embedded; pattern; digital data; detecting; scaling factor"), (ii) the classification tree (G06T1/00, G06T1/0021, G06T1/005, G06T1/0064 "Geometric transform invariant watermarking," H04N1/32144, H04N1/32352), (iii) a Cited By (6) list, and (iv) a Families Citing this family (24) list (only 6 entries rendered). The page as fetched did not include the front-page "References Cited" / examiner-cited list, so I cannot tell you which references were actually of record in the '655 prosecution. I supplemented with search-verified art. Anything I could not verify is flagged [unverified], and no reference below is asserted to have been of-record absent a flag.

Date contradiction to flag: the previously generated sections in this file state today's date is 2026-09-30, while this task states April 26, 2026. I am not resolving it; it does not affect the §103 analysis.

1. The reference date and the §103 field of search

Because the application was filed well before 2013-03-16, pre-AIA §103 governs, and the critical date for printed publications is one year before the US filing (2001-09-11), with the JP priority available under §119 to antedate references falling between 2000-09-12 and 2001-09-11 only for subject matter the JP application supports under §112. That distinction matters a lot here — see §3.3 below, where several superficially on-point Digimarc references have priority dates in December 2000, i.e. after the '655 priority date but before the US filing date.

POSITA. A person of ordinary skill as of September 2000: a B.S. in EE/CS or equivalent plus 2–3 years in image/video signal processing, including working knowledge of (a) transform-domain (DCT/wavelet/Fourier) watermark embedding, (b) standard statistical estimators such as autocorrelation and phase correlation, and (c) MPEG/JPEG block structure. The '655 specification itself is written at approximately that level (it treats autocorrelation as a known tool — "autocorrelation is a method for calculating an optimal period by comparing patterns that themselves have been embedded repetitively").

Claim-construction points that drive the analysis:

• "tuple of base block patterns that are orthogonal to each other" (claims 1, 5, 10, 12, 15, 18, 21, 24, 25) — two vectors whose inner product is zero. This is a broad limitation; orthogonality of basis patterns is standard.
• "has a repetitive pattern in the horizontal and/or vertical direction" — periodicity of the payload-bearing pattern.
• "detecting a scaling factor … from a change of said period" — period estimate → scale = changed period / known period (spec's ρ ≈ o/h, Equation 8 scaling_factor = (c+x_peak)/c).
• "without decompression" (claims 9, 13, 17, 23, 26) — DCT/bitstream-domain detection, no inverse DCT.
• "applying a shift pattern prepared in advance to a detected block" (claim 16) — pre-computed shifted templates used instead of resampling.
• Claim 15's "autocorrelation peak … appearing differently when finding a scaling factor" is a result-oriented / property-of-the-art limitation with no structural hook; it is nearly coextensive with "the watermark is periodic in a block-transform domain."

2. Candidate prior art

2.1 Primary references (high confidence on dates and content)

2.2 References on the "Families Citing this family" list — usable only with care

3. Grounds of rejection

3.1 Ground A — Claims 1, 2, 5, 6, 10, 12, 18, 21, 24, 25

Primary: US 5,636,292 (Rhoads '292) in view of Tirkel (DICTA '95), and further in view of the ordinary knowledge of a POSITA regarding autocorrelation-based period estimation.

What each element maps to:

• "Preparing a digital embedded pattern having information to be embedded as a watermark" — Rhoads '292 claims 1, 5, 6, 8 (information signal + calibration pattern, combined into one embedded signal).
• "A tuple of base block patterns that are orthogonal to each other" — Rhoads '292's "plurality of rotationally symmetric component pattern elements" (claim 1); Cox '97 more explicitly uses mutually orthogonal pseudo-random coefficient vectors.
• "Repetitive pattern in the horizontal and/or vertical direction" / "embedded at least twice or more repetitively" — Rhoads '292 claim 14 ("the information signal is encoded repeatedly throughout the encoded output signal"); Tirkel DICTA '95's design criterion (1) requires high in-phase autocorrelation for rows and columns, which is only achievable with a spatially periodic array.
• "Observing the change of its period upon detection to calculate the scaling factor" (claim 5) — Tirkel's autocorrelation criterion supplies the detector; the arithmetic ρ ≈ o/h ("period changed by o") is elementary and, in the scale-estimation context, already the point of Honsinger US 5,835,639.
• "Known repetition rate" (claims 1, 2, 5, 6) — inherent; you cannot embed a designed periodic pattern without knowing its period.
• "Plural frames / one of said frames" (claims 10, 12) — routine for video; Rhoads '292 is a video/image reference, and the '655 spec itself concedes frame-independent detection is optional ("there is no need to use the same embedded pattern between added frames").

Why a POSITA would combine Rhoads '292 with Tirkel: Both are in digital watermarking for copyright management (analogous art, common field of endeavor). Rhoads '292 supplies the purpose (recover scale so the payload can be decoded) and — via claims 6 and 8 — the architecture in which the calibration data is not a separate payload but is carried by / merged with the information signal. Tirkel supplies the detector criterion: design the watermark array so it has a strong in-phase autocorrelation peak along rows and columns, exactly so that shifts/scale can be resolved by correlation. Combining them is the predictable use of a known estimator (autocorrelation) for its known purpose (period/shift estimation) applied to a known class of pattern (periodic 2-D watermark array). Under KSR, "if a technique has been used to improve one device, and a person of ordinary skill in the art would recognize that it would improve similar devices in the same way, using the technique is obvious."

Why the "double duty" premise does not save the claims. The '655 summary asserts as the feature that "it does not need the specific second embedded pattern for geometric compensation." But Rhoads '292 claim 6 already recites calibration data elements that "are also included in said information signal," and claim 8 recites the calibration signal acting as carrier for the information signal. Once the calibration pattern is also the payload carrier, running autocorrelation on it to recover scale is the natural and predictable next step, not an inventive leap. Rhoads '292 also expressly contemplates that "the representation of the information signal … has geometrical properties useful in ascertaining alignment and registration" (claim 9) — i.e., the payload's geometry is usable for registration.

3.2 Ground B — Claims 3, 7 (and 4, 8) — "without requiring signals for geometric compensation"

These are negative limitations about what the method does not need. A negative limitation does not confer patentability where the prior art renders the positive subject matter obvious and the negative is a mere statement of capability. Two independent theories:

1. Rhoads '292 claim 8 embodies exactly the case where the calibration signal is the payload carrier — so an embodiment with no separate geometric-compensation signal is disclosed.
2. Even under Rhoads '292's other embodiments, the recitation is a characterization of the resulting method, and the '655 method itself uses the embedded pattern as its own registration reference — the very thing Rhoads '292 claim 9 describes.

Claims 4 and 8 (article-of-manufacture versions of 3 and 7) rise and fall with them and add only a computer-usable medium, which is conventional.

3.3 Ground C — Claims 9, 13, 17, 23, 26 — "scaled or compressed data … without decompression"

Ground A + Hartung & Girod (SPIE 2952, 1996) and/or Langelaar et al. (Benelux 1997; JVCIR Dec 1998; IEEE SPM Sept 2000).

Hartung & Girod and Langelaar explicitly teach watermark embedding and detection operating on MPEG/DCT compressed video without full decoding. The motivation is stated in the '655 specification itself and is purely economic: "as the processing of inverse transform is expensive, it is necessary to enable detection of scaling factors without inverse transform." Cost-driven elimination of a known, expensive step is a classic obviousness rationale. Note the '655 specification also concedes the equivalence: "processing similar to the case of baseband data is possible if performing inverse transform on DCT data."

Caveat on timing: the Sept 2000 IEEE Signal Processing Magazine survey by Langelaar et al. sits close to the 2000-09-12 priority date; I would rely on the 1998 JVCIR and 1997 Benelux papers for a clean date, and treat the magazine article as cumulative. [I did not verify the exact publication day of the September 2000 issue — flag.]

3.4 Ground D — Claim 14 (MPEG; 8×4 orthogonal base block patterns)

Ground C + Cox et al. (IEEE TIP Dec 1997) + Kuglin & Hines phase-correlation teaching, and the ordinary skill in block-transform design.

The '655 insight behind claim 14 is that if the base pattern is orthogonal and half a DCT block tall (8×4), the inner product of the DCT-transformed pattern with a cyclically shifted version of itself follows cos(2πs/8) / sin(2πs/8) (the patent's Equation 2), so the phase of the correlation output encodes the fractional block misalignment, giving scale resolution finer than the 8-pixel block grid. Each piece is known:

• Orthogonality of watermark basis vectors in the DCT domain — Cox '97.
• Using phase of a correlation result to obtain sub-block / sub-pixel shift — Kuglin & Hines (1975), textbook phase-correlation registration; also standard in motion estimation.
• Two patterns tiling one 8×8 DCT block (8×4 each) — an immediate consequence of the JPEG/MPEG block geometry the POSITA works in every day.
• Cos/sin quadrature for a phase measurement — elementary signal processing.

Motivation: sub-block precision is required to estimate arbitrary scale factors from block-domain data, since block-quantized correlation alone limits you to 8/h precision (the patent says so). The design space here is small and the result is predictable, which is the KSR "obvious to try" posture. Claim 14 is nonetheless the narrowest and strongest independent claim — see §4.

3.5 Ground E — Claim 15 (compressed data; "autocorrelation peak appears differently")

Ground C plus the DCT-domain watermark references. This claim recites a consequence of periodicity in a block-transform domain rather than any structure. If the pattern is periodic and the data is DCT-blocked, differing apparent autocorrelation peaks with misalignment is inherent. Expect a §112(b) indefiniteness attack on this claim in parallel with §103, and expect it to be read onto any DCT-domain periodic watermark.

3.6 Ground F — Claim 16 — "applying a shift pattern prepared in advance to a detected block"

Grounds C/D + template-matching art (Pereira & Pun, SPIE 3657, Jan 1999) + standard motion-estimation/block-matching practice.

The '655 specification is explicit that the shift approach is a substitution: "the processing needed for enlargement or reduction can be replaced with the shift processing, and further the shift patterns that have been prepared in advance can be used if the shift is limited, thereby making the processing much faster." Replacing resampling with a bank of pre-shifted correlation kernels is (a) a well-known "pre-compute the shifted templates" optimization, and (b) exactly what integer/half-pel motion estimation does in MPEG encoders. Motivation: speed and avoidance of interpolation error. Result: predictable.

3.7 Ground G — Claims 18–22, 24–26 (systems, program products, storage media)

These are apparatus/program-product counterparts of Grounds A–F. Systems and machine-readable media claims add no patentable weight beyond their underlying methods (conventional means-plus-function / program-product language; the hardware of FIG. 23 is a generic PC). They fall with the method claims.

4. Where the obviousness case is weakest (the defensible positions)

An honest analysis has to identify the limitations the art does not squarely teach:

1. Claim 14's specific 8×4 orthogonal tuple with the cos/sin inner-product structure. None of the references I could verify discloses that construction. An examiner or challenger must bridge the gap with a rationale that the 8×4 size and quadrature structure are obvious design choices given phase-correlation and DCT block geometry. That is a defensible but not automatic §103 argument. If I were defending, this is the claim I would focus on — and it is independent, so it standing would not save the others.
2. The negative limitations (claims 3, 7, 4, 8). A patent owner will argue Rhoads '292 requires a calibration signal (claim 1) and therefore teaches away from a scheme that dispenses with one, invoking In re Gurley-style "the art could, but would not" reasoning. The counter is Rhoads '292 claim 8 (calibration signal as carrier) plus the KSR observation that the field was racing toward watermark-only synchronization (see the Dec 2000 Digimarc "synchronization/orientation signals conveying payload data" filings). But I should be candid: this is the argument I would expect to be litigated most heavily, and the outcome would turn on claim 8's disclosure being read as an enabling embodiment.
3. Claims 1/10/12's breadth cuts both ways. "Tuple of base block patterns that are orthogonal to each other" is broad, which makes it easy to read on the art (good for a challenger) but also means the art is more likely to disclose it.
4. No verified reference squarely teaches that the repetition period change — as opposed to an FFT/Fourier-Mellin shift or a template match — is the scale estimator, applied to the payload pattern. Tirkel gives you autocorrelation criteria; Rhoads '292 gives you the calibration-in-the-payload architecture; the joining step (autocorrelate the payload's own period → scale) is supplied by general knowledge rather than by one verified reference. Challengers should either (a) locate a reference that does this explicitly, or (b) rely on KSR's "finite number of identified, predictable solutions" and the fact that autocorrelation-based period estimation is a staple of the POSITA's toolkit. The US 7,657,058 "orientation signals conveying payload data" reference (priority 2000-07-19, i.e. before the '655 priority) is the most likely candidate to close this gap, and I could not retrieve its text — that is the top verification action item.

5. Motivation to combine — consolidated (KSR rationales)

6. Secondary considerations / objective evidence

• No evidence of unexpected results is presented in the specification; the stated advantages (quality, speed, no separate signal) are the expected consequences of removing a second embedded payload, not surprising ones, and the patent does not quantify an unexpected magnitude.
• No confirmed litigation, IPR, or CAFC appeal involving this patent was found (see the Litigation summary in this file). Therefore there is no judicial or PTAB finding of validity or invalidity to cite, and no secondary-considerations record (commercial success, copying, licensing nexus) that I can substantiate. IBM's general watermark licensing program exists, but absent a patent-specific nexus that is not probative.
• Expiration. The record reports the patent expired 2023-08-28 after a 716-day term adjustment. Practical consequence: an IPR/PGR is of little or no value now (no amendment is possible, and any remaining interests would be defensive only), and any past invalidity contention would have had to be raised in a district-court or ITC action — of which I found none.

7. Bottom line

The single most likely-to-succeed §103 case against the core claims is US 5,636,292 in view of Tirkel (DICTA '95), with Hartung & Girod / Langelaar added for the compressed-domain claims and Cox '97 / phase-correlation knowledge added for claim 14. The single most vulnerable claim to survive is claim 14, and the single highest-value unverified item is the content of US 7,657,058 (Digimarc, "Watermark orientation signals conveying payload data," priority 2000-07-19 — before the '655 priority), plus the actual front-page examiner citations for US 7,006,655, which the fetched page did not render.

Verification actions I could not complete (flagged, not asserted): the on-face "References Cited" list of US 7,006,655; the specifications of US 7,657,058, US 6,975,744, US 8,301,453 and US 6,704,431 B1; the exact publication date of the September 2000 IEEE SPM issue; and the §102(e) dates of Digimarc's US 6,408,082 and US 6,424,725 (the latter, "Determining transformations of media signals with embedded code signals," is conceptually very close to the '655 detection claims and should be pulled and date-checked first if it can be established as prior art).