Image Signal Processor AI Patents: Top Companies & Trends 2026
- Filing activity peaked in 2023 at 25 families, then softened toward the 2022 midpoint of 16 — a plateau, not a growth curve, once publication lag is factored in.
- G06T image-processing claims dominate at 75 of 102 families, well ahead of H04N pictorial communication (49) and G06N AI-model claims (40), showing where the claim density actually sits.
- The United States receives 59 filings against 17 for Europe, making the US the primary battleground for freedom-to-operate checks on neural ISP tuning.
Filing growth compares 2021 (14 records) with 2024 (18) — a three-year span. 2024 is the most recent year we treat as complete: publication lags filing by roughly 18 months, so 2025 onwards are still filling in and any growth rate that ends there would understate the field. Top-5 share is the combined record count of the five largest assignees divided by all 102 records in scope (CR5), not by the ranked leaders only.
What this landscape covers
This review tracks patent families filed against image signal processors that use AI-based denoising, scene-adaptive tuning or learned image signal processing in place of, or alongside, fixed-function ISP blocks. The search spans priority years 2015 through the 2026 cut-off, combining classification codes for image data processing (G06T), pictorial communication (H04N) and AI-model computing (G06N) with claim-text signals for machine learning applied to the ISP pipeline.
The 102 families in scope are the fairer unit of analysis than raw publications, since they strip out continuation filings and duplicate national-phase entries that would otherwise inflate any single assignee's apparent footprint.
Let an AI agent run this analysis on your own technology
Pick a task. Every answer cites the patents behind it.
Filing trend and technology composition
Two views matter here: how filing volume has moved year over year, and which IPC subclasses carry the claim weight. Both point to a field that built up quickly, crested in 2023, and is now consolidating around a smaller set of established claim positions rather than expanding into new ones.
A 2023 peak followed by a flattening curve
Filings were effectively zero in 2017 and climbed to a peak of 25 families in 2023. The 2022 midpoint of 16 sits well below that peak, and the partial 2026 count reflects publication lag rather than a real drop-off — but the trajectory from 2022 to 2023 to whatever settles after is flat-to-declining, not a continued ramp.
G06T and H04N carry the claim density
G06T (image data processing) appears in 75 of 102 families and H04N (pictorial communication) in 49, meaning most claims are anchored in image transformation and transmission mechanics rather than in the AI model itself — G06N-classified AI-model claims appear in 40 families, and G06V recognition claims in just 19, suggesting the learning component is more often claimed as an ISP feature than as a standalone model architecture.
Shares are the percentage of the 102 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Go deeper on Image Signal Processor AI and Machine Learning with Eureka
This page is one run against one query. Ask Eureka your own question about image signal processor ai and machine learning and every answer comes back with the patent numbers behind it.
Try EurekaRepresentative and most-cited filings
Predicting optimal values for parameters used in an operation of an image signal processor using machine learning
A method of predicting optimal values for a plurality of parameters used in an operation of an image signal processor includes: inputting initial values for the plurality of parameters to a machine learning model having an input layer, corresponding to the plurality of parameters, and an output layer corresponding to a plurality of evaluation items extracted from a result image generated by the image signal processor; obtaining evaluation scores for the plurality of evaluation items using an output of the machine learning model; adjusting weights, applied to the plurality of parameters, based on the evaluation scores; and determining the optimal values using the adjusted weights.Filed by Samsung Electronics, published 2023-04-20 as US20230117343A1 — a continuation of an earlier family also appearing among the most-cited records in this set.
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US20190108618A1 | Image signal processor for processing images | 111 |
| 2 | US20220164926A1 | Method and device for joint denoising and demosaicing using neural network | 43 |
| 3 | US20200372682A1 | Predicting optimal values for parameters used in an operation of an image signal processor using machine lear… | 36 |
| 4 | US20200211229A1 | Image signal processor for processing images | 34 |
| 5 | US10460231B2 | Method and apparatus of neural network based image signal processor | 33 |
| 6 | US20210105442A1 | Image capture based on action recognition | 20 |
| 7 | US20200389588A1 | Method and system for tuning a camera image signal processor for computer vision tasks | 19 |
| 8 | US20220301123A1 | End to end differentiable machine vision systems, methods, and media | 11 |
| 9 | US10643306B2 | Image signal processor for processing images | 10 |
| 10 | US20250045873A1 | Foveated sensing | 9 |
Citation counts are drawn from a searched corpus and favour older, earlier-filed records; treat them as a signal of influence within this dataset, not as a current-relevance ranking.
Each row carries its publication number; clicking a row searches Eureka by that number.
Put your own technology through the same analysis
Eureka on the web
When you want the answer in the next five minutes.
The agent works the prompt against patents and technical literature, citing every source.
Run your analysis now →MCP server & REST API
When it has to run inside your own pipeline.
Patent search, landscape analysis and assignee resolution as MCP tools. Drop them into any agent framework, or call REST directly.
Browse MCP servers →What the numbers mean for a filing decision
The composition and timing data point to a specific pattern: claim space in the core ISP-plus-ML combination is already dense, while adjacent branches remain comparatively open.
The ramp already happened
Filings rose from near zero in 2017 to a 2023 peak of 25 families, then eased. New entrants filing broad claims on core denoising or auto-tuning methods now compete against an established base rather than an open field.
Claims sit on image processing, not just AI
Nearly three-quarters of families carry a G06T classification against 40 for G06N, meaning most applicants claim the ISP transformation step itself, with the learning model described as a means rather than the invention's core.
The US is the primary filing venue
With 59 of the tracked filings routed through the United States against 17 for Europe, 8 for WIPO/PCT and single digits elsewhere, freedom-to-operate diligence for this technology should start with US prosecution history.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to image signal processor ai and machine learning, with the prior art for and against each one.
Who is filing, and where momentum has stalled
Recent-year momentum data shows a field where the named leaders have gone quiet rather than continuing to build. Several of the most active historical filers show zero filings in the latest tracked year, with year-over-year drops of 100% for at least two of them — consistent with the 2023 peak rather than a still-accelerating race.
Named leaders have gone quiet
Qualcomm, Samsung Electronics, Google, Imagination Technologies and Nvidia all show zero filings in the most recent tracked year, with Samsung and Google each down 100% year over year. This is consistent with a 2023 filing peak followed by a lull, though publication lag means the true 2025-2026 picture is still filling in.
Collaboration is limited and mostly internal
Only five co-assignee pairings appear across the dataset, and the strongest pairing links two units of the same corporate group rather than two independent companies, suggesting most of this technology is developed and claimed in-house.
Filing is concentrated in a handful of offices
Beyond the US and EPO, India, China and the UK each register single-digit filing counts, which narrows the practical scope of a freedom-to-operate search to a small set of jurisdictions for most applicants in this space.
| Assignee | Recent year | YoY |
|---|---|---|
| Qualcomm Incorporated | 0 | — |
| Samsung Electronics Co., Ltd. | 0 | -100% |
| Google LLC | 0 | -100% |
| Imagination Technologies Limited | 0 | — |
| NVIDIA Corporation | 0 | — |
| ALGOLUX INC | 0 | — |
| Intel Corporation | 0 | -100% |
| Qualcomm Technologies, Inc. | 0 | — |
Where to take this next
The dataset points to a field that has already staked out its core claims. The open questions now are about adjacent branches and about which of the quiet leaders re-enters filing once 2024-2026 priority filings finish publishing.
Watch for the lag catching up
Because publication lags filing by roughly 18 months, the apparent 2024-2026 slowdown may partly reverse once pending applications from the named leaders publish. Re-check the trend after the next data refresh before concluding the field is contracting.
Explore filing trends in EurekaMap the white space branches directly
The under-claimed branches identified here are a starting point, not a conclusion. Running a focused search against cross-sensor fusion tuning or on-device model compression claims would confirm how open they actually are.
Run a white space search in EurekaCommon questions about this landscape
The dataset's most-cited records are led by filings naming Samsung Electronics alongside other filings that recur across the ranking, but recent-year momentum data shows several historically active filers, including Qualcomm, Samsung, Google, Imagination Technologies and Nvidia, at zero filings in the latest tracked year. This means historical citation leadership and current filing activity are not the same thing, and a company's ranking in an older citation table should not be read as a sign of present-day activity. Anyone assessing competitive position should look at both the citation table and the recent-year momentum figures together.
Filing activity rose from near zero in 2017 to a peak of 25 families in 2023, then eased toward the 2022 midpoint level of 16, which points to a plateau rather than continued growth. Because publication typically lags filing by around 18 months, the partial counts for the most recent years understate true filing activity and should not yet be read as a decline. The honest read is that the field expanded quickly through 2023 and has since stabilized, with the next one to two years of data needed to confirm the trend.
Most families combine image data processing claims (G06T, present in 75 of 102 families) with pictorial communication claims (H04N, 49 families) and AI-model claims (G06N, 40 families), meaning the typical patent claims an ISP method or pipeline step that happens to use a machine learning model, rather than claiming a novel model architecture on its own. Recognition-specific claims (G06V) appear in only 19 families, and general-purpose computing (G06F) and data recognition (G06K) classifications are minor by comparison. This composition suggests the strongest claim positions sit in the ISP pipeline mechanics, not in the underlying AI model design.
Based on the classification and filing data, the densest claim coverage sits in core denoising, demosaicing and parameter-tuning methods classified under G06T, H04N and G06N. Adjacent areas with thinner coverage in this dataset include cross-sensor fusion tuning for multi-camera systems, on-device model compression for real-time processing, and scene-adaptive tuning applied outside standard visible-spectrum imaging. A prudent next step is running a targeted search against each of these branches rather than assuming thin representation in one broad search equals open white space everywhere.
The United States accounts for 59 of the tracked filings, more than three times the 17 filed at the European Patent Office, with WIPO/PCT, India, China and the United Kingdom each in single digits. This concentration means a freedom-to-operate review focused on the US and, secondarily, Europe will cover the large majority of active filings in this dataset. Companies planning to commercialize in China or India specifically should still run a dedicated local search, since the low counts here may reflect filing strategy rather than an absence of relevant local rights.
Research Image Signal Processor AI and Machine Learning in depth with Eureka
Go past this page: query the whole image signal processor ai and machine learning corpus yourself, in your own scope.
Every answer comes back with patent numbers you can open.
Disclaimer. This page is generated from Patsnap Eureka data drawn from a limited snapshot of global patent and scientific-literature records, and is provided for general information and reference only.
Patent data carries inherent limitations: recent filings (typically the most recent 18–24 months) are under-counted due to standard publication lag; counts may be reported at either a patent-family or a patent-record basis and are not always directly comparable; classification, applicant-name, and citation data may contain errors, duplicates, or omissions; and the underlying search query defines and constrains the scope shown. As a result, the analysis may be incomplete or inaccurate and may not reflect the full technology landscape.
Nothing on this page constitutes an exhaustive prior-art, novelty, freedom-to-operate, or validity search, nor does it constitute legal, financial, investment, or professional advice, and it should not be relied upon as such. Any patent, commercial, or strategic decision should be verified independently and reviewed with qualified patent, legal, and domain professionals. Patsnap makes no warranties, express or implied, as to the accuracy, completeness, or fitness for any particular purpose of the information presented.
Machine translation. Assignee and organisation names originally recorded in Chinese, Japanese or Korean have been rendered into English by an AI translation step so that the tables stay readable. These renderings are best-effort and may not match a company’s registered English name; the original name is what the underlying patent record carries, and it is what any Eureka query launched from this page uses.