Picture Imperfect: Providing Improved Resolution For Analogous-Art

The Nielsen Co. (US), LLC v. TVision Insights, Inc., No. 25-1371 (Fed. Cir. Aug. 14, 2026)

U.S. Patent No. 11,470,243 (“the ‘243 patent”) covers audience measurement systems (essentially the technology that watches you while you watch television) ostensibly to count how many people are in the room and whether they are paying attention. The ‘243 patent describes a camera capturing images of the “media exposure environment,” and processor circuitry that performs two operations on those images:

• A “people counter” detects human features (heads, faces) using, according to the specification, low-resolution images.
• A “person identifier” performs facial recognition against a database of facial signatures, which the specification says requires high-resolution images.

The specification devotes a significant amount of discussion to a specific engineering problem: frequent activation of illumination sources that is needed for the high-resolution facial-recognition step drains power, shortens the illumination source’s life, generates heat, and (in a detail with unexpected comedic value in the resulting litigation) annoys the people being measured.

The main claims at issue recite reducing the resolution of a first image and determining head orientation from that reduced-resolution image. Additional claims added a two-step process: generating a facial signature from a region of a second image corresponding to the head location, then comparing this second image against a database of facial signatures. Notably, none of the challenged claims recited an illumination source at all. That omission turned out to matter a great deal.

Before getting into the procedural history and the court’s reasoning, here are the practical takeaways for practitioners drafting and prosecuting similar applications:

Best Practice Tips for Preparation and Prosecution

1. Match the problem statement to the claim scope. If the specification highlights a narrow, specific problem (here, illumination-driven power drain and heat), but the claims are drafted broadly enough to omit any reference to the mechanism causing that problem, expect the broader claim language to define the relevant art. In this case, a vivid backstory about a light bulb will not save claims that never mention one.
2. If a specific mechanism is the actual differentiator, claim it. The specification’s binning technique may have been the inventive contribution, but claim 4 recited generic resolution reduction. Claiming the specific mechanism (or including it in at least one independent claim) preserves the narrower scope if broad claims fall.
3. Do not count on courts reading preferred embodiments into claims. The court declined to import binning, first-image exclusivity, or full-resolution second-image requirements from the specification. If a limitation matters, put it in the claim.
4. Build a technical-effect record during prosecution, not just in the background section. If a claimed architecture produces a specific, non-obvious result, develop that record (declarations, comparative data) rather than relying on the examiner or the Board inferring it from a generalized efficiency rationale. Be careful with this one, however, as it cuts both ways and can just as easily support an obviousness combination.
5. Broaden prior-art searches beyond the nominal commercial field. Tian came from expression-analysis research, not audience measurement, and was still analogous art because it addressed the same underlying technical problem (facial detection/analysis at varying resolutions). Search by technical task, not by target market.
6. When attacking a combination based on analogous art, avoid conclusory arguments (“different fields,” “no efficiency problem identified in the primary reference”). These will likely be found insufficient, as they were here. A stronger response identifies materially different technical problems between the references, technical incompatibilities or tradeoffs undermining the combination, or evidence that the asserted benefit would not actually result from the claimed combination. If needed, you can support these by declarations or data where the benefit depends on factual assumptions rather than common engineering knowledge.
7. Preserve narrower fallback claims tied to the actual disclosed mechanism. The two illumination-source dependent claims (24–25) were never challenged. A narrow claim anchored to the specific structure disclosed can be the one still standing after the broader claims are gone. This is a point worth remembering the next time claim drafting feels like an exercise in maximizing scope for its own sake.

Procedural History

TVision Insights filed a petition for inter partes review of the ‘243 patent challenging fourteen claims. Nielsen disclaimed the three challenged independent claims in its preliminary response and the Board instituted review on claims 4–6, 8, 11–14, and 18–20.

TVision asserted two obviousness grounds, both anchored to the same secondary reference: a scientific publication by Ying-li Tian titled “Evaluation of Face Resolution for Expression Analysis,” which studied how well facial expression analysis performs on down-sampled (lower-resolution) images.

The Board held that Tian was analogous art (resting on the “reasonably pertinent” prong without needing to reach field-of-endeavor) and finding all challenged claims unpatentable under both obviousness grounds. Nielsen appealed to the Federal Circuit, disputing (1) whether Tian was analogous art at all, and (2) whether Tian’s disclosures rendered the claims obvious even if analogous.

Patent Preparation/Prosecution Issues Highlighted by the Case

The ‘243 patent is a study in the gap between what a specification emphasizes and what the claims actually cover:

• The problem stated in the specification is narrow: illumination-driven power drain, heat, shortened component life, and audience annoyance. The claims, however, say nothing about illumination and are written at the level of generic image processing, facial detection, and facial analysis. The Federal Circuit treated the claims as defining the relevant problem for analogous-art purposes.
• Claim drafting also left the resolution-reduction limitation broad. The specification describes a specific technique (pixel “binning,” which trades resolution for contrast), but the base claim in the IPR (claim 4) recites only reducing the resolution, with no requirement of binning, that only the first image in a sequence be reduced, or that the second image (used for facial-signature generation in claims 5–6) be full-resolution. Nielsen argued for all three limitations as a matter of claim construction; the court found none of them in the actual language in the claim and declined to import them from the specification.
• Two dependent claims (24–25) did recite an illumination source and were not challenged.

The throughline: Nielsen told a specific, sympathetic story about illumination fatigue in the specification, then claimed a solution broad enough to sweep in reference material about facial expression analysis that had nothing to do with light sources or the problem at issue. However, the court held that the claims, not the narrative, decide the analogous-art inquiry.

CAFC Decision and Reasoning

• Analogous Art – Notice/APA Argument: Nielsen argued the Board violated the APA by finding Tian “reasonably pertinent” when TVision’s petition argued only that Tian was in the same field of endeavor. The panel rejected this on three independent grounds: (1) the field-of-endeavor and reasonably-pertinent evidence overlapped, since the same expert testimony about Tian’s disclosure of “head detection and pose estimation” supported both prongs; (2) an analogous-art theory need not be stated with precision and may be implicit in the petition’s discussion of the claims and motivation to combine; and (3) Nielsen conceded it identified no evidence or argument it was deprived of the chance to present, so any error was harmless under the APA’s prejudicial-error rule.
• Analogous Art – Merits: The court held that the relevant “problem” for analogous-art purposes is not confined to whatever drawback the specification singles out; a patent can address multiple problems, and the objective reach of the claims (plus the disclosure as a whole) controls. Because the challenged claims recite image processing, facial detection, and facial analysis, with no illumination limitation, the panel held that the Board reasonably characterized the problem as one of image processing and facial detection generally. Tian, which studies facial expression analysis performance across image resolutions, addressed that same problem and therefore “logically would have commended itself” to the inventor’s attention, regardless of Tian’s different commercial field (expression analysis versus audience measurement).
• Obviousness: Nielsen tried to read the binning technique, a “first-image-only” resolution reduction, and a full-resolution second image into claim 4 (and its dependents). The CAFC declined each request: the claim language did not support any of these limitations, so Tian’s disclosure of down-sampling (a different resolution-reduction technique) sufficed, and Tian’s disclosure of detecting a face in a first frame and tracking it through subsequent frames satisfied the two-step facial-recognition limitation of claims 5–6.
• Motivation to Combine: The CAFC upheld the Board finding of motivation to combine (which was supported by expert testimony that reducing resolution saves processing resources and time while preserving adequate detection performance) was substantial evidence, consistent with precedent recognizing that generalized efficiency rationales can supply motivation to combine even without an explicit textual cue in the references themselves. In other words, “this would probably run faster” remains a viable, if unglamorous, motivation-to-combine theory.

The panel thus affirmed the Board’s unpatentability determination in full.

This article is provided for general informational purposes and does not constitute legal advice.