US 4977521

I'll verify against external sources before summarizing.

Let me check for any litigation or docket records involving this patent.

Preliminary note on the searches

I searched for the exact identifier 4977521 and for litigation/docket references tied to it. The bibliographic and text sources returned (Google Patents, Justia, SumoBrain, the EPO Global Patent Index entry for the family member EP 0426712 A1, and EPO/ISA search reports citing "US 4 977 521 A (M.C. KAPLAN) 11 December 1990") agree in all material respects with the patent text you supplied. I found no record of any lawsuit, PTAB proceeding, or CAFC docket involving US 4977521; the "4977521" strings that surfaced in litigation-style searches were unrelated internal case IDs in Chinese and Brazilian court-system documents and one unrelated PDF citation number. I cannot prove a negative from a keyword search, so treat "no litigation found" as no such record surfaced, not as a certified docket clearance. I also did not obtain a live USPTO Patent Center maintenance-fee ledger; status below is from the Google Patents legal-status field.

I did not auto-correct any identifier. 4977521 corresponds to US 4,977,521, application US 07/223,812.

Bibliographic summary

Abstract (as given in the patent)

At least one frame of a photographic film strip is exposed to an image; before development, six (or so) calibration patches in a specially reserved frame on the same film are individually exposed to six predetermined calibration levels spanning the film's exposure latitude. One film layer is developed to carry a positive image in one color and a negative image in another color using positive and negative dyes. The film is scanned to determine the positive and negative dye amounts D_p and D_n at each scanner pixel. The likeliest exposure value E as a function of D_n and D_p is determined probabilistically from six histograms obtained by scanning the six calibration patches; all such E values are stored in a look-up table addressed by the corresponding D_p, D_n values. Images on the film are then scanned pixel-by-pixel, each pixel's D_p, D_n addressing the look-up table to obtain its likeliest exposure value, which is sent to a printer. Combining the positive and negative images reduces film noise so as to obtain the likeliest exposure value at each pixel with the smallest error possible, with no prior knowledge of film characteristics.

Plain-language overview of the independent claims

There are four independent claims: 1 and 21 (method/system, generic), and 42 and 47 (system, implemented with two-color development and a look-up table).

Claim 1 — Method (generic Bayesian read-out).
Obtain, for a film, K distributions of paired latent and non-latent grain counts (D_p, D_n) at K known exposure values. From those K distributions, precompute a likeliest exposure value E_i for each of a range of possible (D_p, D_n) pairs, producing a table of likeliest exposure values. Then measure the (D_p, D_n) pair at an actual pixel in the exposed film frame and pick the matching precomputed value, building up a set of likeliest exposure values that reconstructs the recorded image.
Note on labels: in claim 1 D_p is stated to be the latent grain count and D_n the non-latent grain count — the reverse of the specification's usage, where D_p is the positive dye from non-latent grains and D_n the negative dye from latent grains. The same flipped mapping appears in claim 42/47's "first color"/"second color" language. I flag this as an apparent labeling inconsistency in the claims, not as a correction.

Claim 21 — System (generic apparatus counterpart of claim 1).
Means for developing latent and non-latent grains so each count can be measured separately; means for storing the K distributions of (D_p, D_n) for K predetermined exposures; means for computing the likeliest exposure E_i for each of many grain-count pairs; and means for measuring a pair at each pixel and selecting the corresponding stored likeliest value to produce an enhanced image.

Claim 42 — System with two-color development, scanner, processors, and look-up table.
Latent-grain developer that lays down a dye of a first color at each latent grain, plus a non-latent-grain developer that lays down a dye of a second color at each non-latent grain, together producing a two-colored developed image. A scanner views that image as an array of N pixels and measures the first-color dye amount D_p and second-color dye amount D_n (per calibration patch). A first processor turns the scanner data into K distributions P(D_n, D_p | E_i) (one "cloud" per calibration patch). A second processor computes, for the range of joint dye counts, a likeliest exposure E as a normalized weighted sum of each distribution times its corresponding exposure level E_i. A look-up table stores E against each (D_n, D_p) address. A third processor reads the table for every pixel of the user image, accumulating an array of exposure values that constitutes the enhanced image.

Claim 47 — System, essentially the same pipeline as claim 42 stated differently.
Same two-color latent/non-latent development and scanner, but the first processor is defined as determining the probability distribution P(D_n, D_p | E_i) for each exposure level; the second processor determines likeliest exposure E as a normalized weighted sum over the distributions; a look-up table stores E per joint dye count; and means are provided for addressing that table with each sensed pixel's (D_n, D_p) to retrieve the likeliest exposure and build the enhanced image.

What the dependents add (briefly)

• Claims 2–3, 22–24, 26: the calibration patches are exposed on the same film strip/roll as the picture frame and developed simultaneously, so the K distributions are "exactly characteristic" of that film; a single developing means serves both.
• Claims 4, 25: printing an image from the recovered exposure values.
• Claim 5: developing the user frame so latent and non-latent counts are separately measurable.
• Claims 6–9, 27–30: the interpolation refinement — define N exposure values spanning the K calibration exposures, derive N distributions, and read off the likeliest exposure from those.
• Claims 8 and 29: the specific interpolation mechanics — compute mean and variance of D_n, mean and variance of D_p, and the covariance of D_n with D_p for each of the K distributions; interpolate those five magnitudes at N points (the description specifies quasi-cubic Hermite spline interpolation); then construct an interpolated Gaussian (multivariate normal) distribution at each of the N points from those five values.
• Claims 10, 31: N = 2^m, where each exposure value is an m-bit word (e.g., 256 levels for 8-bit data).
• Claims 11–15, 32–36, 46, 51: weight each distribution by an a priori exposure probability distribution P(E) (derived by histogramming a large archive of representative photographs — consumer images or astronomical images, depending on the application); the weighting is expressed as a Bayes-theorem normalized weighted sum.
• Claims 16–18, 37–39, 45, 50: scratch/defect detection and suppression — flag a pixel as defective when the probability of its (D_n, D_p) is below a threshold across all exposure levels, skip the table lookup for it, and substitute a value derived from neighboring pixels (the specification describes growing a neighborhood until a non-defective pixel is found and substituting the median).
• Claims 19–20, 40–41: sequential development with a different color dye for latent vs. non-latent grains, one developed with the positive dye and the other with the negative dye.
• Claims 43–44, 48–49: the processor-level implementation of the interpolation and normalized weighted summation (the "##EQU4##" / "##EQU5##" summation forms) used in the claim-42/47 systems.

Technical summary and stated advantage

The patent attacks film granularity — random pixel-to-pixel variation in silver halide grain counts — by developing a single emulsion layer twice in complementary modes: some latent grains are developed to a negative cyan dye (D_n), the remaining latent grains are developed black-and-white, and all non-latent grains are then developed to a positive magenta dye (D_p). A CCD scanner reads each calibration patch through a red filter (cyan complement) and a green filter (magenta complement) to build, per exposure level, a joint histogram of (D_p, D_n) — treated as the conditional distribution P(D_p, D_n | E_i). Bayes' theorem inverts this into P(E_i | D_p, D_n), and the mean exposure E is the exposure-weighted sum normalized by the sum of the probabilities (equations presented as ##EQU1## through ##EQU4## / ##EQU5##, with the a priori term dropped in the simplified version, ##EQU3##). The result is precomputed for all 256² joint density values and stored in a look-up table, so user-frame processing is just table addressing.

The patent expressly positions itself against the Bird reference, which optimizes detective quantum efficiency but assumes an idealized monodisperse emulsion. The claimed approach requires no model of the film at all — no assumptions about grain size distribution, grains per pixel, fog, stain, developer nonlinearity, or grain photon thresholds — because all noise sources (grain density/size variation, absorption cross-section, photon thresholds, grain obscuration, post-development dye cloud effects, and scanner noise) are absorbed empirically into the calibration-patch distributions. The specification asserts the result is the minimum-mean-squared-error estimate and therefore "optimal" in that sense.

Experimental data are reported from a real single-layer emulsion, with calibration exposures separated by 0.3 log exposure using a standard neutral-density step tablet; FIGS. 8a–8d show measured "clouds," FIGS. 9a–9d show 15 interpolated multivariate-Gaussian clouds (the actual run used 256), and FIGS. 10a–10d compare the regenerated clouds to the measured ones.

Confidence and uncertainty

• High confidence on title, inventor, assignee, application number, filing date (1988-07-25), issue date (1990-12-11), claim count (51), and the family members — these are consistent across the patent text and the independent sources found.
• Moderate confidence on "no litigation": no such record appeared in my searches, but I could not query USPTO Patent Center or CAFC dockets directly, so I cannot rule out a proceeding that is not indexed in general web results.
• Flagged ambiguity (my own reading, not a correction): the D_p/D_n grain-type mapping in claims 1 and 42/47 appears inverted relative to the specification's D_p = positive (non-latent) dye, D_n = negative (latent) dye convention. Anyone relying on claim scope should read the claims as issued rather than harmonizing them with the description.

I'll search for litigation involving US patent 4977521 across patent litigation databases.

Let me run more targeted searches on patent litigation databases.

Let me do a few more targeted searches to be thorough.

Litigation Search Report — U.S. Patent No. 4,977,521

Result: No litigation found

I searched patent-litigation sources (including queries aimed at Unified Patents' litigation portal, Docket Alarm, CourtListener/RECAP, RPX, and general CAFC/PACER-indexed web content) for the specific number 4,977,521. I found no record of any civil action, ITC investigation, PTAB proceeding, or CAFC appeal involving U.S. Patent No. 4,977,521. I cannot supply a plaintiff, defendant, jurisdiction, case number, filing date, or outcome because no case matching this patent number appeared in any result.

What the patent is (for identification purposes only)

• Number: US 4,977,521 (do not confuse with the different patent 7,259,521, which appears in unrelated PTAB material and belongs to a Kodak v. CTP Innovations matter — that is a different patent).
• Title: "Film noise reduction by application of Bayes theorem to positive/negative film"
• Inventor: Martin C. Kaplan
• Original/current assignee: Eastman Kodak Company
• Application: US 07/223,812; filed 1988‑07‑25; granted 1990‑12‑11
• Legal status: Expired – Lifetime (anticipated expiration 2008‑07‑25)
• Foreign family (PCT route): WO1990001240A1, EP0426712A1, JPH04500587A

What the searches did surface (and why none of it is litigation against this patent)

1. Google Patents record — lists no "Litigation" or "Reexamination" events, only citation data (28 third‑party documents citing it, and a list of patents in the family citing it). Citations are prior‑art references, not lawsuits.
2. EPO/other search reports — US 4,977,521 appears as an "A" (background) reference in the search reports of EP 1,453,299 and EP 0,916,995, and in the NL 1007441 PCT search report. Being cited as prior art is not litigation.
3. False‑positive numeric matches — several unrelated items contained the digit string "4977521" without being the patent:
   • Chinese court filing IDs "4977521f…" / "4977521d…" (2024–2025 docket entries, no patent content);
   • A journal DOI "10.1063/1.4977521" (a chemistry article);
   • Lithuanian business‑registry notice numbers (e.g., "Pranešimo Nr. 4977521");
   • A Brazilian court enrollment list entry "4977521‑9".
     None of these concerns U.S. Patent 4,977,521.

Confidence and caveats

• Confidence: moderately high that no litigation exists, based on the absence of any hits across multiple litigation‑oriented queries and the fact that the patent is a 1990 Eastman Kodak film‑processing patent that expired in 2008 and does not appear in the "Cited By"/litigation cross‑references.
• Limitation: I was unable to open the Unified Patents litigation caselist or a Docket Alarm/CourtListener docket page directly for this number in this session; my conclusion rests on the search results returned. Older pre‑2000 district‑court files and settled matters are sometimes not fully indexed in free databases, so a residual possibility of an un‑indexed or pre‑digital case cannot be entirely excluded.
• To be fully certain, the authoritative next steps would be: (a) a direct PACER docket search on the patent number; (b) the Unified Patents portal caselist filtered by patent 4,977,521; and (c) the USPTO Patent Center "Litigation" / assignment tabs for application 07/223,812.

If no litigation is your expected answer, that is consistent with everything I found: there is no known litigation involving U.S. Patent No. 4,977,521.

PTAB / AIA Trial Assessment — U.S. Patent No. 4,977,521

Patent: Film noise reduction by application of Bayes theorem to positive/negative film
Assignee of record: Eastman Kodak Company (original assignee and current assignee per Google Patents)
Inventor: Martin C. Kaplan
Application: US 07/223,812 · Filed: 1988-07-25 · Granted: 1990-12-11
Legal status on file: Expired – Lifetime; anticipated expiration 2008-07-25
Family: EP 0426712 A1, JP JPH04500587A, WO 1990001240 A1 (all sharing the 1988-07-25 priority date)

Proceedings overview

Total AIA trial proceedings on file: 0. Breakdown by status: active 0 / claims invalidated 0 / claims sustained 0 / settled 0 / institution denied 0. The USPTO Open Data Portal structured "PTAB proceedings on file" block for US 4,977,521 returns no IPR, PGR, or CBM proceedings, and independent web searches for IPR/PGR/CBM petitions naming this patent surfaced none (see note on incidental PTACTS hits below). The bottom-line defensive posture for a defendant today is not "hardened patent" and not "invalidated patent" — it is an expired patent with no litigated validity record at all: no PTAB panel has ever construed a claim, no Final Written Decision exists, and no § 315(e)(2) estoppel has attached to anyone. The real defensive question is not whether claims 1–51 survive IPR (they were never tested) but whether the patent was enforceable at all during the relevant damages period. On the face of the record, it expired 2008-07-25, roughly four years before the first IPR was ever filed at the PTAB (AIA trials became available 2012-09-16), which structurally explains the empty docket.

No AIA trial proceedings to report

There are no proceedings to describe at claim-level granularity, and I will not invent proceeding numbers, panels, or dispositions. For completeness, here is what I affirmatively checked:

• ODP structured data (canonical source): no AIA trial proceedings indexed for US 4,977,521.
• Board-side record: Google Patents' litigation/trial linkage for US 4,977,521 (https://patents.google.com/patent/US4977521/en) shows no PTAB proceeding entries, only a "Cited By (28)" list of later patents. Linkage between a patent and its PTAB proceedings is imperfect, so absence there is weak evidence on its own.
• Search-based sweep: queries targeting "4977521" IPR, "4977521" "inter partes review" / "covered business method" / "post-grant review" petition, and litigation-oriented queries returned nothing tying this patent to any AIA trial. The only ptacts.uspto.gov results that surfaced were unrelated documents (petitions quoting Kodak's Bayer patent EP 0152434, and CBM/§ 101 papers citing third-party patents) that matched on "Eastman Kodak" or on unrelated claim text — none references US 4,977,521 as the challenged patent. I am flagging these explicitly so the negative result is auditable rather than assumed.

Why the docket is empty (structural, not accidental):

1. Term. 20-year term from the 1988-07-25 filing ran to 2008-07-25 per the record. AIA trials did not exist until 2012-09-16. The entire IPR/PGR/CBM window opened only after the patent's term ended.
2. PGR ineligible. Post-grant review under 35 U.S.C. § 321 applies only to patents with claims having an effective filing date on or after 2013-03-16. This patent's effective filing date is 1988-07-25.
3. CBM ineligible in substance. CBM review (AIA § 18) covered patents claiming a method/apparatus for performing data processing or other operations used in the practice, administration, or management of a financial product or service, excluding technological inventions. This patent claims photographic film development, Bayesian exposure estimation, look-up tables, and scratch suppression — a technological invention, not a financial one. The CBM program also sunset 2020-09-16.
4. IPR was theoretically possible post-2008 but economically inert. The Board has instituted IPRs against expired patents in limited circumstances, but a petitioner must still have a concrete stake (e.g., an outstanding infringement suit under § 315(b)). No such suit appears in the sources searched for this patent, and none would be expected for a film-emulsion processing patent whose commercial era ended with digital capture.

Incidental note on the citation web (not a proceeding). US 4,977,521 is cited by 28 later publications, including Applied Science Fiction's electronic film development family (e.g., US 6,441,871; US 6,437,358) and later Kodak film-grain work (e.g., US 5,641,596). That makes it a prior-art reference of record for other people's patents — a role it has played since 1990 — but it cuts the other way from a PTAB standpoint: the patent was never itself put on trial.

Strategic summary

Claim status. No claim of US 4,977,521 is CANCELED, SUSTAINED, or even UNTESTED-in-the-IPR-sense—because no IPR was ever filed. All 51 claims (methods claims 1–20, system claims 21–41, and system claims 42–51) stood as issued from 1990-12-11 until the patent expired on 2008-07-25. Nothing was narrowed by reissue, reexamination, disclaimer, or certificate of correction on the records I searched (I could not exhaustively rule out a § 1.321 statutory disclaimer from the sources available; the Google Patents claims listing shows all 51 claims intact). There is no surviving-claims list to give a defendant, and equally no dead-claim list to attack a demand letter with.

Estoppel landscape. There is no § 315(e)(2) estoppel against anyone, because there was no petitioner and no final written decision. That is a double-edged fact: a defendant gains no free pass from a prior petitioner's forfeited grounds, but also faces no precedent finding claims valid. If this patent were somehow asserted today, prior-art defenses would be litigated on a blank slate — and the primary § 102/§ 103 reference would likely be the Bird article the patent itself disparages (Normal Development, Reversal Development, and Composite Processing, Photographic Science and Engineering, Vol. 22, No. 6, pp. 328–335, Nov./Dec. 1978), which the specification identifies as the closest prior art and distinguishes only on the ground that Bird assumes a monodisperse emulsion. That is the same art the examiner already considered, so an IPR would have turned on whether the claim-1 "K distributions … representative of said film" limitation, and the claim-6/8 interpolation limitations, read on Bird alone or on Bird in view of the admitted calibrations arts cited in the family's foreign search reports (e.g., GB 1,173,565; FR 2,602,596).

Pattern signals. No repeat-petitioner pattern (zero petitioners). No patent-owner PTAB appeal history (no FWD to appeal, and no Federal Circuit docket for this patent surfaced). No defensive aggregator involvement — no Unified Patents, RPX, or similar filing appears anywhere in the record, which is consistent with the patent's 2008 expiration predating the defensive-aggregator IPR era. The Federal Circuit docket shows no appeal attributable to this patent.

Recommended next steps

If you are a defendant and received a demand citing this patent:

1. Lead with expiration. Open with the face-of-the-record term end date, 2008-07-25 ("Expired – Lifetime"; anticipated expiration 2008-07-25 per the Google Patents bibliographic record at https://patents.google.com/patent/US4977521/en). A patent that expired roughly eighteen years ago cannot support a claim for ongoing or future infringement, and pre-expiration damages exposure for film-emulsion processing conduct is almost certainly time-barred under 35 U.S.C. § 286 (six-year limit) even before laches is considered. This is a stronger and simpler position than any invalidity theory.
2. There is no FWD to link. Do not expect a "claim 1 is dead" citation — no Final Written Decision exists for this patent. Any representation by a counterparty that this patent "survived PTAB" or "was upheld by the Board" is unsupported by the record and should be challenged directly. Conversely, do not represent to a court that any claim was canceled.
3. Verify the record yourself before relying on this memo. Confirm on USPTO Patent Center / PTAB E2E (https://ptab.uspto.gov) and via the ODP API that no proceeding, reexamination, or disclaimer exists, and confirm the maintenance-fee/expiration status. My sources are the ODP structured block, Google Patents, and open web searches; I could not inspect a certified Patent Center file history, so treat the "zero proceedings" finding as high-confidence but verify-able rather than certified.
4. If any live assertion is found, the trial-stage milestones to monitor would be: § 315(b) one-year bar from service of the complaint; institution decision due within six months of the petition's filing date (§ 314(b)); and the § 316(a)(11) statutory final-written-decision deadline of one year from institution (extendable up to six months for good cause). None of these clocks are running now.
5. Absence of IPRs is itself a signal here — and in this case it points at expiration, not at strength. Well-asserted, economically significant patents attract IPRs. This one attracted none, because by the time the AIA trial regime existed the patent had already expired and its subject matter (silver-halide emulsion noise reduction) had been superseded commercially. Treat the empty docket as corroboration of an unmonetizable patent, not as evidence of an untested-but-valid estate.

Caveats. I found no PTAB proceeding, and I likewise found no district court litigation or Federal Circuit appeal involving this patent in the sources searched — but I cannot prove a universal negative from open search. If you have a complaint, a demand letter, or an asserted-patent list in hand, provide it and I will run the specific proceeding-number and docket checks against PTAB E2E and CourtListener (https://www.courtlistener.com) rather than relying on patent-number searches alone.

I'll start by checking the live Assignment Center / USPTO records and related sources before writing anything.

Let me try to pull the Assignment Center record directly and check for any post-issuance transfer.

Assignment & Ownership Analysis — US 4,977,521

Patent: Film noise reduction by application of bayes theorem to positive/negative film
App. No.: US 07/223,812 · Filed: 1988-07-25 · Issued: 1990-12-11 · Anticipated expiration: 2008-07-25 · Status: Expired – Lifetime

Source limitation, stated up front: I was unable to load the Assignment Center/PEDS record page directly. The ownership facts below come from the Google Patents legal-events block for US4977521 (which mirrors USPTO assignment data) and secondary sources. Where the material I have does not disclose a reel/frame or correspondent, I say so rather than supplying a number. Verify at the Assignment Center search page: https://assignmentcenter.uspto.gov/ (mirror: https://assignment.uspto.gov/patent/index.html).

Inventors

• No adverse / unusual pattern. There is a single inventor, and he did not depart Kodak around the filing. Publication records attribute Kaplan to Eastman Kodak across his career (a later affiliation with Carestream Health appears — the Kodak Health Group spin-off sold to Onex in 2007 — but that is ~19 years post-filing and unrelated to this patent's chain).
• No "all inventors left the assignee within 12 months of filing" signal here; that pattern simply doesn't apply to a one-inventor, stayed-put fact pattern.

Original assignee

Eastman Kodak Company, a New Jersey corporation, 343 State Street, Rochester, NY 14650 (assignee of record and current assignee per Google Patents).

• Primary line of business (at issue date): silver-halide photographic film, photofinishing, and image capture — the exact domain of the claims (dual latent/non-latent dye development plus digital scanning and a Bayes-theorem look-up table).
• Did Kodak ship a product embodying the claims? Unclear. The disclosed process (partial latent-grain developer → B&W developer → non-latent developer, two-color positive/negative image, CCD scan, LUT) is an R&D/processing pipeline, not a discrete commercial article. I found no evidence Kodak ever commercialized this specific positive/negative dual-dye development workflow. Treat "shipped a product" as unverified.
• Current status: Operating (restructured). Kodak filed Chapter 11 on 2012-01-19 in the U.S. Bankruptcy Court, S.D.N.Y., and emerged ~September 2013 as a commercial-imaging company (SEC Form 8-K/press release, 2012-12-19). It has not dissolved.
• Foreign family (counterparts, not assignments): WO1990001240A1 (1990-02-08), EP0426712A1 (1991-05-15), JPH04500587A / JP1508096A (PCT filed 1989-07-17). These are priority filings, not separate ownership events.

Assignment timeline

One (1) recorded assignment exists for this patent. No post-issuance conveyances are on record.

• ~1988 (executed; exact execution date not shown) / recorded 1990-02-21 — Reel/Frame not disclosed in the sources I can access (Google Patents shows the event date only; the '521 grant was a pre-1990s-film reel era, but I will not invent a reel/frame).
  • Conveyance: Assignment — "ASSIGNMENT OF ASSIGNORS INTEREST"
  • Assignor: Martin C. Kaplan
  • Assignee: Eastman Kodak Company, a Corp. of NJ
  • Correspondent: Not disclosed in the material available to me. (Assignment recordings filed by Kodak in this era were customarily handled by Kodak's own Patent Department, 343 State Street, Rochester, NY — flagging this as unverified, not a finding.)
  • Context: Original inventor-to-employer assignment; this is the instrument that vested title in Kodak. Nothing more.

No other assignment records. Absent any later conveyance, the chain stops at the original assignee: Kodak remains assignee of record, and the patent expired 2008-07-25. No Security Agreement, no Merger, no Change of Name, no Release tied to this patent number appears in the record.

Important non-finding to pre-empt a common error: Kodak's December 2012 sale of ~1,100 digital-imaging patents to Intellectual Ventures (buyer) with RPX Corporation (organizer) and 12 licensees (Apple, Google, Microsoft, RIM, Samsung, Adobe, HTC, Facebook, Fujifilm, Huawei, Amazon, Shutterfly) is a real event — but US 4,977,521 expired on 2008-07-25, roughly four years earlier. An expired patent has no assertion value and I found no evidence it was included in that portfolio. Do not attribute the 2012 IV/RPX transaction to this patent without a reel/frame showing it.

Timeline diagram

timeline
    title Ownership of US 4977521
    1988 : Application filed by Kaplan
         : Assigned to Eastman Kodak
    1990 : Patent issued to Kodak
    2008 : Patent expired
    2012 : Kodak files Chapter 11
         : 521 already expired in 2008

NPE / troll-pattern signals

1. Shell-entity transfer — not present. The only conveyance is inventor → Eastman Kodak (recorded 1990-02-21). No "IP / Patents / Licensing / Holdings / Ventures" entity ever appears in the chain for this patent number.

2. Known asserter in the chain — not present. No Acacia, Marathon, Intellectual Ventures, IPNav, Wi-LAN, Mosaid/Conversant, Vringo, Pendrell, Innovatio, MPHJ, Lumen View, Round Rock, Document Generation, or Spangenberg entity is an assignee of record. (IV is the general buyer of Kodak's 2012 imaging portfolio, but no IV assignment to this patent is recorded — see the non-finding above.)

3. Repeat correspondent across the chain — unclear. The single recorded assignment does not expose a correspondent in the material I can access, so there is nothing to test for recurrence. No repeat-player attorney to flag.

4. Cascading transfers — not present. Zero post-issuance assignments. No chained LLCs, hence no shared-correspondent or shared-address pattern.

5. Pre-litigation transfer — not present. No assignment within 6 months of any suit; no infringement litigation naming this patent appears in the sources reviewed, and the patent expired in 2008.

6. Bankruptcy fire-sale — not present (for this patent). Kodak did file Chapter 11 (2012-01-19) and sell ~1,100 imaging patents to IV/RPX (announced 2012-12-19), but '521 expired 2008-07-25 — before the bankruptcy — so it could not have been part of a going-concern assertion package and is not evidenced in the sale.

7. Privateering — not present. No operating-company → NPE transfer of this patent exists on record.

8. Defensive aggregator (anti-NPE) — not present. The chain does not terminate at RPX, AST, LOT, Unified Patents, or OIN. (RPX's 2012 role was in the general Kodak portfolio, not this expired patent.)

Verdict

Insufficient data — only the original inventor-to-employer assignment is on record.

The sole recorded conveyance is the 1990-02-21 assignment of Martin C. Kaplan's interest to Eastman Kodak Company; there are no post-issuance transfers, and Kodak remains assignee of record. The patent expired 2008-07-25, roughly four years before Kodak's 2012 Chapter 11 and the IV/RPX sale of its digital-imaging portfolio, so there is no NPE chain to reconstruct and no basis to attribute the 2012 transaction to this patent. If a fresh Assignment Center pull shows a reel/frame beyond the original assignment, that would change this call — until then, this is a dead, unmonetized Kodak patent.

Verification links:

• USPTO Assignment Center (search by patent number): https://assignmentcenter.uspto.gov/
• Mirror: https://assignment.uspto.gov/patent/index.html
• Google Patents legal events: https://patents.google.com/patent/US4977521/en
• Kodak Ch. 11 patent sale (context only): SEC Form 8-K Ex. 99.1, 2012-12-19

I'll search the USPTO/public patent databases for the specific number 4977521 and pull its cited prior-art references.

I have the citation list. Let me verify each cited reference individually.

Now the remaining cited references:

I hit the tool-call limit mid-verification, so here is the analysis built on what I retrieved. I flag unverified items explicitly rather than filling gaps from memory.

Prior-Art Analysis — US 4,977,521 (application 07/223,812)

Search provenance and a source conflict to flag first

Identifiers interpreted literally: "4977521" = US 4,977,521, application US 07/223,812, "Film noise reduction by application of Bayes theorem to positive/negative film," Eastman Kodak, filed 1988-07-25, issued 1990-12-11. No other number is substituted.

Provenance issue worth flagging: the full patent text supplied to me for this analysis (the Google Patents scrape) contained the "Cited By" list (28 documents) but not the front-page "References Cited" block. I therefore reconstructed the cited-art list from (a) a SumoBrain reproduction of the printed front page of US 4,977,521, and (b) Google Patents' own citation count. This is a genuine gap in the primary source and should be noted if claim-scope or validity work depends on the exact list; the authoritative artifact is the printed patent front page (or the PDF at the Google Patents "Download PDF" link).

Cross-check: Google Patents' US 4,977,521 record carries "Patent Citations (7)" — matching the seven U.S. patents on the SumoBrain front-page reproduction. No discrepancy found.

On "most relevant": I found no record of a reexamination, IPR, or validity contest, so there is no adjudicated "most relevant prior art" ranking. My ranking below is an examiner-style assessment, not a holding.

The cited references of record

Note on item 5: the front-page listing reads "4777102 … Levine." I did not independently verify that five-digit-plus-check rendering in a second source before running out of searches, so treat the last digit with normal diligence — the intended reference is a 4,777,xxx Levine patent.

Legal sequencing point (pre-AIA §102 governs here, filing 1988-07-25): references 5–7 published after the July 25, 1988 filing date. They cannot be §102(a) or §102(b) art. They are available only as §102(e) art, and only if their respective U.S. filing dates antedate Kaplan's date of invention. That is almost certainly why the examiner reached for them — they are data-processing/quantitation art, not film-response art.

Reference-by-reference analysis

1. US 4,090,243 — Kotera et al., "Color separating method and apparatus using statistical techniques" (Matsushita Electric Industrial Co.), issued May 16, 1978

Verified. This is the closest of the seven to the statistical heart of the patent, and the most interesting citation on the face of the document.

What it discloses: color separation of a painted print by statistical classification. A microscopic scanner and a macroscopic color separator generate spectral component signals. The system computes the a priori probability of occurrence P(Cᵢ) of each of n color samples, and the conditional probability density p(x|Cᵢ) that an observed spectral component corresponds to color sample Cᵢ, and then identifies the observed dot as color Cᵢ by maximizing the joint quantity P(Cᵢ)·p(x|Cᵢ) — i.e., a maximum-a-posteriori / Bayes-style decision rule. It even computes P(Cᵢ) from an inverse-matrix operation on mean spectral reflectivities.

§102 assessment: Does not anticipate any claim. On its face it lacks, for claim 1/21: latent and non-latent grain count pairs D_p, D_n; calibration patches at K predetermined exposure values on the film; and a likeliest exposure value as the output. And for claims 42/47: the two-color latent/non-latent development, the look-up table addressed by joint dye amounts, and the normalized weighted sum over exposure values Eᵢ.
Closest claims: claims 13–15, 32–36, 46, 51 (weighting distributions by an a priori probability distribution, expressed as a Bayes-theorem normalized weighted sum), and claims 11/12/33/35 conceptually. Kotera is best characterized as category "A" background, with a plausible §103 role against the a priori-probability dependent claims — the patent's own specification concedes that harvesting P(E) by histogramming photographs is conventional.

2. US 4,573,493 — Blackwell et al., "Gray scale image processor," issued March 4, 1986

Verified. Monochrome-camera inspection system. Each picture element is assigned a gray-scale value; elements above a threshold are assumed to correspond to the object of interest; the gray values are collected into a histogram of frequency of occurrence; extreme values are discarded; the remainder are averaged to determine a nominal gray scale, which is compared against predetermined nominal values.

§102 assessment: Does not anticipate any claim. No latent/non-latent grain counts, no positive/negative dye pairing, no exposure estimation, no Bayes inversion.
Closest claims: claim 2/22 and claim 6/27 tangentially — Blackwell does build a distribution of pixel counts versus a measured value, which is structurally analogous to the calibration-patch histogram of FIG. 2. That is the only overlap, and it is a building block, not the claimed combination. Category "A."

3. US 4,638,456 — Elias et al., "Spot quantitation," issued January 20, 1987

Verified (title and subject confirmed; related to copending application Ser. No. 574,713, filed Jan. 27, 1984). Method and system for improving the ability to quantitate the amount of flux or material in localized collections — i.e., "spots."

§102 assessment: Does not anticipate any claim. Relevant only as general art for the measurement of a quantity within a defined patch, which maps loosely onto the calibration-patch measurement in claims 2/22.
Closest claims: claims 2, 22 (measuring grain-count pairs in each discrete pixel in K calibration patches). Category "A."

4. US 4,666,307 — Matsumoto et al., issued May 19, 1987

NOT VERIFIED in this session. I obtained the number, date, and inventor name from the front-page reproduction but did not retrieve the title, abstract, or disclosure before my search budget was exhausted. I will not guess its subject matter. Recommendation: pull the PDF at patents.google.com/patent/US4666307 and read the abstract before relying on any §102 characterization. On its date alone it is pre-filing art available under §102(a)/(b).

5. US 4,777,102 — Levine, issued Oct. 11, 1988

NOT VERIFIED. Number/date/inventor from the front-page listing only. Post-filing date → §102(e) only. Do not characterize without reading the document.

6. US 4,788,131 — Kellogg et al., issued Nov. 29, 1988

NOT VERIFIED. Number/date/inventors from the front-page listing only. Post-filing → §102(e) only.

7. US 4,812,390 — Giannesi, issued Mar. 14, 1989

NOT VERIFIED. Number/date/inventor from the front-page listing only. Post-filing → §102(e) only.

The most relevant prior art is not a patent: Bird (1978)

Full citation: Bird, "Normal Development, Reversal Development, and Composite Processing: A New Method for Gaining a Simultaneous Improvement in Latitude and Detective Quantum Efficiency in Silver Halide Films," Photographic Science and Engineering, Vol. 22, No. 6, pp. 328–335 (Nov./Dec. 1978). Available under §102(b) as a printed publication more than one year before the 1988-07-25 filing.

This is the only reference the specification engages on the merits, and it is the only cited item that discloses the core physical premise of the invention: independently developing latent and non-latent silver halide grains and combining the two resulting densities to improve latitude and detective quantum efficiency.

§102 assessment: Does not anticipate the independent claims (1, 21, 42, 47), but comes closest. The specification itself supplies the reason: Bird "proposes an algorithm for combining the densities of D_n and D_p by individually weighting them and then computing the exposure from the combination," but does so "requiring certain assumptions in film characteristics, the main assumption being an idealized monodisperse emulsion." Bird therefore lacks:

• the K calibration patches at predetermined exposure levels on the same film (claims 2, 3, 22–24);
• the empirical K distributions P(D_p, D_n | Eᵢ) derived from scanning those patches (claims 2, 42, 47);
• the Bayesian inversion to a likeliest exposure E (claims 1, 15, 42, 47);
• the look-up table addressed by joint dye amount (claims 42, 47).

Closest claims: claims 5, 19, 20, 23, 40, 41 — the two-color, sequential latent/non-latent development steps. These are the claims most exposed to a Bird-based §102 or §103 attack, because the patent's own Background section concedes Bird teaches joint positive/negative development of latent and non-latent grains. The patent's escape hatch is that the independent claims all require the calibration-patch/Bayes machinery that Bird does not have.

Bottom line

Most relevant prior art overall: the Bird 1978 article, cited-and-distinguished in the Background — not any of the seven U.S. patents.

Anticipation (35 U.S.C. §102) conclusion: On the material available to me, no single cited reference anticipates any of the four independent claims (1, 21, 42, 47), or any dependent claim. Every cited U.S. patent is directed to a different problem — colorimetric classification (Kotera), industrial gray-scale inspection (Blackwell), spot quantitation (Elias), and three §102(e)-only data-processing references — and none discloses the combination of (i) same-film calibration patches, (ii) separately-measured latent/non-latent grain or dye counts, (iii) empirical per-exposure distributions of the joint pair, and (iv) a Bayesian likeliest-exposure readout via look-up table. The patents functioned as category "A" background art supporting the examiner's finding of a complete disclosure, not as anticipatory art.

Where a challenger would actually push: §103, not §102 — combining Kotera (a priori probability × conditional density, i.e., the Bayes maximum) with Bird (joint positive/negative grain development), and adding Blackwell/Elias for the histogram-and-patch measurement mechanics. The patent fortifies itself against exactly that attack: it argues most of the probabilistic steps were known and claims the specific empirical, model-free pipeline plus the interpolation (claims 8, 29, 43–44, 48–49) and scratch detection (claims 16–18, 37–39, 45, 50). The claims least insulated from a §103 combination are the generic method/system claims 1 and 21 read without their interpolation and a-priori-probability dependents.

Confidence and open items

• High confidence: the identity of US 4,977,521 and its status; the seven-reference list and dates as reproduced from the printed front page; the full disclosure of Kotera, Blackwell, Elias, and Bird; the pre-AIA §102 sequencing analysis of references 5–7.
• Unverified (do not rely on without a PDF pull): the subject matter of US 4,666,307, 4,777,102, 4,788,131, 4,812,390. I retrieved numbers and dates only. My §102 characterization of these four is deliberately withheld rather than inferred.
• Source limitation flagged above: the primary patent text I was given omitted the front-page "References Cited" block; the list is reconstructed from a secondary reproduction. Verify against the printed patent before filing anything that depends on it.
• Cross-reference to earlier sections: the previously generated "Patent summary" identified Bird as the notable prior art of record. That remains correct and is now supplemented with the seven U.S. patents; I found no contradiction between the earlier sections and this search. The earlier D_p/D_n claim-labeling flag (claims 1 and 42/47 using the inverse mapping from the specification) also matters for §102/§103 work: a reference that discloses non-latent→positive dye and latent→negative dye does read on claim 1's "latent grain count D_p," so a challenger can argue the flipped labels are mere nomenclature. That is a scope argument, not an anticipation finding.

I'll ground this obviousness analysis in actual prior-art searches before assembling the § 103 combinations.

Let me pin down the specific prior-art references of record and the closest non-patent art.

Obviousness Analysis Under 35 U.S.C. § 103 — U.S. Patent No. 4,977,521

Building on the claim mapping and technical summary already generated. This section assumes that prior work and does not repeat it.

1. Framing and the four Graham factors

A § 103 challenge is evaluated as of the critical date, 1988‑07‑25 (the application's filing date, which is also the earliest priority date; the PCT/EP/JP filings were 1989‑07‑17 and add nothing). I apply the Graham v. John Deere factors — scope/content of the prior art, differences from the claims, and level of ordinary skill — with the KSR guidance that a combination of known elements is obvious where it yields no more than predictable results, and the TriVascular caution against hindsight reconstruction without an articulated reason to combine.

Hypothetical person having ordinary skill in the art (PHOSITA). I would define this as a person with a bachelor's degree in physics, electrical engineering, or photographic science and roughly 3–5 years' experience in one or both of (a) silver halide sensitometry / photofinishing, and (b) digital image processing and applied statistics. This is a hybrid art, and the claims sit squarely on the seam between the two — which matters, because the strongest obviousness grounds are the ones that pair a film‑physics reference with a statistical‑estimation reference. (I state this as my construct; the patent file contains no express PHOSITA definition, so a challenger's expert declaration would be the actual vehicle for it.)

Analogous art. Both arts are reasonably pertinent: the film‑development and two‑channel‑scanning references address the very problem the patent states ("film noise attributable to ... random variations in the density of active grains"), and the Bayesian/MAP image‑restoration references address the very inference method the claims use.

A caveat before I proceed. § 103 obviousness is a legal conclusion; I am assembling the prima facie technical case and the counter‑case, not rendering judgment. I also could not retrieve the disclosures of several U.S. references of record (noted in § 2), so those are treated as unverified and are not load‑bearing below.

2. The prior‑art record actually available

Important note on the applicant's own admissions. The specification itself concedes what Bird teaches: "The Bird reference proposes an algorithm for combining the densities of D_n and D_p by individually weighting them and then computing the exposure from the combination." Applicant admissions about the prior art are strong evidence of its scope. That admission supplies, from Bird alone: dual development, separate measurement of D_n and D_p, and computing an exposure from a weighted combination of the two.

3. Mapping the independent claims; where the gaps are

Claim 1 (method) breaks into four steps:

1. (a) obtain K distributions of (D_p, D_n) pairs for K predetermined exposure values, representative of the film;
2. (b) from those distributions, determine a likeliest exposure E for each of a plurality of (D_p, D_n) pairs, generating a table of precomputed values;
3. (c) measure the (D_p, D_n) pair at an actual pixel and select the corresponding precomputed likeliest value;
4. (d) accumulate the selected values into a set representing the image.

Bird supplies the two‑channel D_n/D_p acquisition and the combining‑to‑exposure concept (steps a and b in embryo), but it does not disclose (i) forming empirical distributions of joint (D_p, D_n) from calibration exposures, (ii) applying Bayes' theorem to invert P(D_p,D_n|E) into P(E|D_p,D_n), or (iii) precomputing and storing a per‑pair lookup. Those are the differences that any ground must close.

4. Grounds of rejection (combinations + motivations)

Ground 1 — Bird in view of Kotera (both on the face of the patent)

Combination: Bird's dual‑development/two‑channel scan plus Kotera's Bayes‑rule statistical estimator with Gaussian conditional densities and a priori probabilities.

What each supplies and how the gap closes:

• Bird → element (a)'s acquisition: two physically distinct channels measured on the same film, and the express goal of "computing the exposure from the combination."
• Kotera → elements (b) and (c): the exact computation P(C_i)·p(x|C_i) in which P(C_i) is an a priori probability and p(x|C_i) is a conditional density; evaluating the product for each candidate class and selecting the best; and storing the pre‑computed conditional‑density data in a storage device so each scanned point is processed against stored statistics. Substituting exposure levels E_i for Kotera's color classes C_i, and joint (D_p, D_n) for Kotera's spectral vector x, applies Kotera's method to Bird's data by direct substitution.
• Kotera also independently supplies the multivariate‑Gaussian‑with‑covariance machinery that underlies claims 8, 29, 44 and 49 (the interpolated Gaussian "clouds"), including inverse covariance, determinant, and the log form.

Motivation to combine (articulated): Both references are drawn from the same practical field — scanning a film/photographic record to recover a physically meaningful quantity per pixel, using digitized multi‑channel signals — and both confront the same pathology (ambiguous, noisy measured values that make deterministic inversion unreliable). Kotera states its estimator is an "unbiased estimator ... used for any color pattern as long as the same color samples are employed," i.e., it is presented as a general‑purpose statistical estimation tool, not one tethered to colorimetry. A PHOSITA seeking to improve on Bird's linear weighted mixing — which Bird itself obtains by assuming a monodisperse emulsion — would look to the standard statistical estimator that (i) uses the same two channels, (ii) requires no parametric film model, and (iii) is taught as reusable. The references are analogous art, the combination is a substitution of one known estimation technique for another in the same data pipeline, and the result (per‑pixel exposure estimates) is exactly the predictable output each reference seeks. That is the KSR "arrangement of old elements" situation.

Predicted output / reasonable expectation of success: High. Kotera's construction is explicitly designed to be run in real time against a store of pre‑computed statistics as the scanner advances — functionally the look‑up‑table architecture of claims 1(c) and 42/47.

Ground 2 — Bird + a Bayesian image‑restoration reference (Hunt 1977 / Geman & Geman 1984; optionally Tavildar 1985) + a sensitometric calibration reference

Combination: Bird (dual positive/negative development and two‑channel scanning) in view of a MAP/Bayes image‑restoration reference, further in view of the conventional sensitometric step‑wedge/calibration‑patch practice.

Why this closes the gaps differently from Ground 1: This ground attacks the "distributions from calibration exposures" and "Bayes inversion" elements head‑on rather than via Kotera's classifier framing:

• Hunt (1977) is the closest art on the central insight. By the reference listing, Hunt "explicitly includes nonlinear relations between intensity and film density, by use of the D‑log E curve, and derives a maximum a posteriori (Bayes) estimate of the restored image." That is a Bayesian inversion of a film's density response to recover the underlying signal — the same statistical problem the claims solve, in the same physical medium.
• Geman & Geman (1984) supplies the general theorem that the MAP estimate under a posterior defined by a degradation model plus a prior gives the "most probable states under the [posterior] distribution," including multiplicative or additive noise — the framework the claims use.
• Tavildar (c. 1985) applies MAP estimation specifically to film grain noise, removing any argument that the Bayesian approach was confined to non‑photographic imaging.
• Sensitometric calibration — exposing a film strip to a graded series of known exposures and measuring the resulting densities — is the foundational technique of the art (the patent itself describes its six patches as spanning the film latitude and its real‑film experiment uses a "standard granularity neutral density step tablet"). A PHOSITA would regard calibrating a Bayesian estimator on the same film as the natural way to obtain the prior/conditional statistics, and WO 87/02852 (Kodak) shows the same assignee already applying histogram‑based statistics to scanned color film.

Motivation to combine: The estimator is only as good as its estimate of the conditional density P(D_p,D_n|E), and the only way to obtain that density without assuming a film model (which Bird's approach requires and the specification disclaims) is to measure it. Measured calibration references are the standard tool for exactly that. The combination produces the predictable benefit of a model‑free Bayesian estimator.

Ground 3 — Grounds 1/2 + Powell (1982)

Combination: Either of the above, further in view of Powell et al., "A method for the digital enhancement of unsharp, grainy photographic images" (Proc. Int'l Conf. on Electronic Image Processing, 1982), which appears on the PCT search report for this family.

Relevance: Powell confirms that, by 1982, the problem of digitally enhancing grainy photographic images was itself a recognized field, with the noise‑from‑grain problem and its digital treatment already framed. That supplies the "known problem, known solution" predicate of KSR and rebuts any argument that combining a digital statistical estimator with a film‑grain problem was an unpredictable leap. It also strengthens the case for the "enhanced image output" elements (claims 4, 25).

Ground 4 — Any of the above + known defect‑detection/median‑filter art (claims 16–18, 37–39, 45, 50)

Combination: Any primary ground above, in view of the well‑established image‑defect art. The patent's own dependent claims characterize the scratch element narrowly: flag a pixel when its measured (D_n, D_p) yields low probability, skip the lookup, and substitute a value derived from neighbors (the spec's median‑of‑a‑grown neighborhood).

Relevance and motivation: Once the Bayes processor is computing a per‑pixel probability, "flag the improbable pixels and replace them from their neighbors" is the textbook use of an outlier detector plus a median/rank‑order filter — a technique that long pre‑dated 1988 and was applied to scanned‑image spot/scratch defects. The patent itself concedes the point: "The particular algorithm used for scratch suppression is not so important as the fact that scratches may be detected." That concession frames the scratch claims as the application of a known repair technique to a signal the novel estimator happens to make available. Caveat: I did not verify the specific disclosures of defect‑repair art in this search, so a challenger would need to supply a concrete reference (e.g., a scan‑defect repair patent or a rank‑order‑filter text) with a verified pre‑1988 date.

Ground 5 — KSR "design choice / obvious to try" treatment of the narrow dependents

Several dependents add little beyond predictable engineering selections and, on their own or with the grounds above, are vulnerable:

5. The strongest arguments against obviousness

A challenger should expect these, and I want to state them plainly because they are the reason I would not characterize the patent as clearly invalid on this record:

1. Teaching away / "unworkable prior approach." The specification does not merely improve on Bird — it disparages the entire modelling paradigm Bird exemplifies ("the prior approach ... was unworkable"; Bird "optimizes the detective quantum efficiency ... requiring ... an idealized monodisperse emulsion"). Its stated invention is to eschew any model and let the calibration patches supply an empirical "understanding." Where the primary reference's method is described as resting on a false assumption, a PHOSITA could argue there was no reason to use Bird's framework at all — which undercuts the "modify Bird" motivation. This is a genuine, citable counter‑argument, though under KSR "a reference's mere preference" is not necessarily a teaching away, and Bird still does disclose the two‑channel combination the claims require.

2. Change in principle, not degree. Bird's optimum mixing is a linear weighting derived from a monodisperse model. The claims compute a posterior distribution and a Bayes‑optimal estimate, which the specification asserts is provably minimum‑mean‑squared‑error for real (polydisperse) film. A patent owner would argue this is a different kind of solution, not a better‑tuned version of the same one, and that the empirical‑distribution step (not in Bird, not in Kotera's color setting) is the inventive core.

3. The joint, same‑film, on‑film calibration concept. Requiring the (D_p, D_n) distributions to be measured on the very film being processed — via patches on the customer's roll — is the element least clearly present in any single reference I located, and it is what the specification leans on for the "no prior knowledge" advantage. Absent a reference teaching same‑strip calibration for this purpose, this is the claim element most likely to survive.

4. No secondary considerations run in the applicant's favor — but none run against it either. There is no evidence of record of commercial success, a long‑felt need, or industry praise, and no litigation (see the earlier litigation section). That means nothing rebuts a prima facie case, but it also means a challenger gets no help from the patentee's litigation history.

5. Reference‑verification gaps. To make Grounds 1/2/4 litigation‑grade, a challenger must (a) obtain and verify the actual disclosures of Bird and Powell, (b) resolve the § 102(e) filing dates of Levine, Kellogg and Giannesi, and (c) supply a concrete, dated defect‑repair reference for Ground 4. My searches confirmed the existence and bibliographic data of these references but not all of their technical substance; I flag that as a real evidentiary limitation, not a formality.

6. Bottom line

• The most likely‑to‑succeed combination is Bird (1978) + Kotera US 4,090,243, because both are references of record, both are analogous art, and together they disclose essentially every element of independent claim 1: dual positive/negative acquisition (Bird), the (D_p,D_n) → exposure combination (Bird's admitted "optimum mixing"), and the Bayes‑rule P(·)·p(·|·) computation with a priori weighting, a multivariate‑Gaussian conditional density, pre‑computed storage, and per‑pixel selection (Kotera). Kotera also supplies the Gaussian/covariance machinery recited in the interpolation dependents. The articulated motivation is KSR-style substitution of a known, expressly general‑purpose statistical estimator for a known, model‑dependent linear mixer operating on the same two‑channel data.
• A strong secondary combination is Bird + Hunt (1977) / Geman & Geman (1984) [optionally Tavildar (1985)] + conventional sensitometric calibration [optionally Powell (1982)], which attacks the empirical‑distribution and Bayes‑inversion elements directly rather than through a classifier analogy.
• The hardest claim elements to invalidate are the empirical joint (D_p,D_n) distributions measured on the same film strip as the image frame, and the recasting of the problem from model‑based linear mixing to model‑free Bayes‑optimal estimation of real, polydisperse film. Those are where a patent owner would concentrate, and I would not predict an outcome either way on this record.
• I am not rendering a legal conclusion of obviousness. The above is the technical prima facie case and its rebuttals; the ultimate question is for a court or the PTAB on a full evidentiary record that I do not have.

Sources cited: https://patents.google.com/patent/US4977521/en (specification, admissions as to Bird, claims) · https://patents.justia.com/patent/4977521 (list of cited U.S. references) · https://patents.google.com/patent/WO1990001240A1/en (PCT search‑report citations: DE 3402823, WO 8702852, Bird, Powell) · https://patents.searchlight.law/doc/US4090243 and the USPTO PDF of US 4,090,243 (Kotera; Bayes' rule, a priori P(C_i), multivariate normal with covariance, stored statistics) · https://ieeexplore.ieee.org/document/4767596 (Geman & Geman 1984, MAP restoration) · 1985 "Picture processing" survey listing Tavildar, "Maximum a posteriori estimation in presence of film grain noise," Signal Processing 8:363.