Fastest
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 →https://www.patsnap.com/resources/blog/rd-blog/cmos-image-sensor-ai-and-machine-learning-patent-landscape/ · Patsnap · data cut-off 2026-07-31 · downloaded from the live page
Sensors & MEMS · Patent LandscapeCMOS Image Sensor AI and Machine Learning Patents
- Filing activity peaked in 2023 at 10 records then eased back toward the 2022 midpoint of 6, suggesting the field is past its first filing wave rather than still accelerating.
- H04N pictorial-communication claims dominate appearing in 47 of 59 records, while G06N AI-model claims appear in only 10 — most inventions still frame the sensor, not the model, as the point of novelty.
- The United States receives the largest share of filings with 25 records, ahead of PCT filings at 14 and EPO at 10, pointing to a US-centred first-filing strategy for this niche.
59
Published Records
58%
Top-5 Share of All Records
+150%
3-Yr Growth (lag-adjusted)
US
Leading Jurisdiction
Published byPatsnap Research··7 min readSourced from Patsnap Eureka
Overview
What this landscape covers
This review tracks patent families at the intersection of CMOS image sensors and machine learning: pixel-level or near-sensor computation, computational imaging pipelines, and on-chip neural network inference applied to imaging. The underlying search combines title/abstract terms for image sensor and in-sensor computing with IPC classifications spanning H04N25 (image sensor circuitry), G06N3 (neural network computing) and H01L27/146 (photosensitive semiconductor devices), so it captures both the hardware substrate and the algorithmic layer sitting on top of it.
Coverage runs from 2015 through the 2026 data cut-off, though the most recent one to two years are understated because publication typically lags filing by around 18 months. With 59 total patent families, this is a compact, still-forming niche rather than a mature, high-volume field — read the rankings below as a map of who has staked early claims, not a settled hierarchy.
Filing activity by year, 2017–2026
Source: Patsnap Eureka. Assignee ranking and totals. Derived from a Patsnap search on CMOS Image Sensor AI and Machine Learning covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
Let an AI agent run this analysis on your own technology
Pick a task. Every answer cites the patents behind it.
10,000 free credits to start
The data
Filing trend and technology composition
Two views of the same 59 families: how filing volume has moved year over year, and which IPC subclasses carry the claim weight.
A filing curve that has already crested
Annual filings rose from 3 in 2017 to a peak of 10 in 2023, then pulled back toward the 2022 level of 6. Combined with the publication lag on 2025-2026 filings, the honest reading is a field whose first wave of patenting has already happened, with the second wave yet to show up in the public record.
Sensor circuitry outweighs AI-model claims
H04N (pictorial communication) appears in 47 of 59 records and H01L (semiconductor devices) in 19, against just 10 for G06N (AI-model computing) and 9 for G06T (image data processing). Most applicants are still patenting the sensor architecture and readout path, with the learned model treated as an attached feature rather than the primary inventive concept.
Shares are the percentage of the 59 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Source: Patsnap Eureka. Filing trend and technology composition. Derived from a Patsnap search on CMOS Image Sensor AI and Machine Learning covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.
Go deeper on CMOS Image Sensor AI and Machine Learning with Eureka
This page is one run against one query. Ask Eureka your own question about cmos image sensor ai and machine learning and every answer comes back with the patent numbers behind it.
Try EurekaKey patents
Representative filing and most-cited prior art
Representative record
WO2022225962A1 — Photon-counting CMOS image sensor for time-of-flight ranging
TRUSTEES OF DARTMOUTH COLLEGE
A photon-counting CMOS quanta image sensor can include a photon-counting pixel and a binary readout circuitry. The photon-counting pixel can be configured to generate an output voltage change when impinged by a photon. The binary readout circuitry can be coupled to the photon-counting pixel, and can be configured to output a first binary signal when receiving the output voltage change and a second binary signal when not receiving the output voltage change.Filed by Trustees of Dartmouth College, published 2022-10-27, this claims a binary-readout photon-counting pixel architecture aimed at time-of-flight ranging rather than conventional intensity imaging.
Most-cited records in this corpus
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US20160309065A1 | Light guided image plane tiled arrays with dense fiber optic bundles for light-field and high resolution imag… | 112 |
| 2 | US20170139131A1 | Coherent fiber array with dense fiber optic bundles for light-field and high resolution image acquisition | 41 |
| 3 | US20160006913A1 | Optical imaging apparatus, in particular for computational imaging, having further functionality | 27 |
| 4 | US20220206434A1 | System and method for deep learning-based color holographic microscopy | 18 |
| 5 | US20130063622A1 | Image sensor and method of capturing an image | 13 |
| 6 | US20220341782A1 | Image sensor and preparation method thereof, and electronic device | 11 |
| 7 | WO2018201219A1 | CMOS image sensors with pixel-wise programmable exposure encoding and methods for use of same | 11 |
| 8 | US10229943B2 | Method and system for pixel-wise imaging | 10 |
| 9 | WO2018195669A1 | Method and system for pixel-wise imaging | 9 |
| 10 | US20200351466A1 | Low Power Framework for Controlling Image Sensor Mode in a Mobile Image Capture Device | 8 |
Citation counts accumulate over time, so older filings such as the light-field and fiber-bundle imaging patents lead the table by influence rather than by current commercial relevance.
Each row carries its publication number; clicking a row searches Eureka by that number.
Source: Patsnap Eureka. Citation counts and representative records. Derived from a Patsnap search on CMOS Image Sensor AI and Machine Learning covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
Run it yourself
Put your own technology through the same analysis
Where to run it
For builders
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 →Insights
What the numbers mean for a filing decision
Three read-throughs from the trend, classification and citation data above.
Filing momentum
10 in 2023
peak year
The wave has crested, not started
Filings climbed from 3 in 2017 to a peak of 10 in 2023 before easing back to the 2022 midpoint of 6. Given the 18-month publication lag, 2025-2026 numbers will fill in somewhat, but the shape of the curve already looks post-peak rather than early-stage.
Based on 59 patent families, 2017-2026
Claim distribution
47 vs 10
H04N vs G06N records
Sensor hardware still carries the novelty
H04N pictorial-communication claims appear in 47 of 59 records; G06N AI-model claims appear in only 10. Novelty is overwhelmingly anchored in sensor circuitry and readout design, with the neural network typically claimed as an application rather than the invention itself.
IPC subclass counts, this corpus
Filing geography
25 US filings
largest receiving office
A US-first filing pattern with thin multi-jurisdiction cover
The United States receives 25 filings, ahead of PCT applications at 14 and the EPO at 10, with Canada and China at 3 each. That gap between US filings and PCT/EPO cover suggests many applicants are not yet pursuing broad international protection for this specific claim combination.
Receiving office counts, this corpus
Citation concentration
112 citations
top-cited record
Influence sits with older imaging-hardware patents
The most-cited record in the corpus, on light-guided tiled image-plane arrays, carries 112 citations — far ahead of the newest deep-learning-based entries. That gap reflects age and corpus search terms more than present-day relevance; recent AI-processing filings simply have not had time to accumulate citations.
Citation counts, most-cited records
Eureka AI Agent
Looking for what nobody has claimed yet?
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to cmos image sensor ai and machine learning, with the prior art for and against each one.
Source: Patsnap Eureka. Co-assignee relationships and derived observations. Derived from a Patsnap search on CMOS Image Sensor AI and Machine Learning covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
Players
Who is filing, and where the field is still open
Assignee activity in this corpus is thin rather than concentrated: several named organisations show zero filings in the latest tracked year, consistent with a niche that has passed its first filing peak. Co-assignee pairs are rare — only 6 across the whole set — meaning most work here is being filed by single organisations rather than through joint ventures.
Academic filers
0 latest-year filings
University of Toronto, Dartmouth
Universities anchor the early record but have gone quiet recently
University of Toronto and the Trustees of Dartmouth College both appear in this dataset, including the representative photon-counting sensor record, but neither shows activity in the most recent tracked year.
Recent-year momentum table
Established imaging suppliers
0 latest-year, -100% YoY
Sony Semiconductor Solutions
A named sensor leader has pulled back sharply
Sony Semiconductor Solutions shows a -100% year-on-year change with zero filings in the latest tracked year, despite Sony-affiliated entities appearing among the corpus's co-assignee pairs. That is a pause worth watching rather than an exit signal, given publication lag.
Recent-year momentum table
Platform and optics entrants
0 latest-year filings
Huawei, Google, ams Osram (Heptagon)
Large platform players are present but not currently active
Huawei, Google and ams-affiliated Heptagon Micro Optics all appear in the assignee set with zero filings in the latest tracked year. Their presence at all signals this claim space is on the radar of large players, even where current filing volume is low.
Recent-year momentum table
Recent-year filing momentum by assignee
| Assignee | Recent year | YoY |
|---|---|---|
| University of Toronto | 0 | — |
| Huawei Technologies Co., Ltd. | 0 | — |
| Dream Chip Technologies GmbH | 0 | — |
| Sony Semiconductor Solutions Corporation | 0 | -100% |
| Heptagon Micro Optics Pte Ltd (ams Osram) | 0 | — |
| Google LLC | 0 | — |
| Lytro, Inc. | 0 | — |
| Sony Europe B.V. | 0 | -100% |
Source: Patsnap Eureka. Assignee-level momentum. Derived from a Patsnap search on CMOS Image Sensor AI and Machine Learning covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
What's next
Where to take this analysis
The dataset points to open ground rather than a settled ranking. Two directions make sense next.
Map the readout-architecture sub-claims in detail
With H04N claims at 47 of 59 records and G06N claims at only 10, the sensor readout layer is where most freedom-to-operate risk actually sits, not the AI model layer most teams assume is contested.
Explore in EurekaTrack dormant assignees for re-entry signals
Several named organisations, including Sony Semiconductor Solutions, Huawei and Google, show zero filings in the latest tracked year. A renewed filing from any of them would be a meaningful signal given how thin current activity is.
Set up monitoring in EurekaSource: Patsnap Eureka. Forward-looking reading of the same dataset. Derived from a Patsnap search on CMOS Image Sensor AI and Machine Learning covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
FAQ
Common questions about this landscape
This dataset contains 59 published patent families matching the combined search for CMOS image sensor terms and in-sensor computing, computational imaging, on-chip neural network or AI image processing terms, filtered to the relevant IPC classes. That is a compact figure for a technology area, meaning individual filings carry more weight than they would in a crowded field. Coverage runs from 2015 through the 2026 cut-off, with the most recent one to two years understated due to publication lag.
Filing activity rose from 3 records in 2017 to a peak of 10 in 2023, then eased back toward the 2022 midpoint of 6. That pattern reads as a field past its first filing wave rather than one still accelerating, though the newest years will fill in somewhat once lagging publications appear. Anyone benchmarking momentum should treat 2025-2026 figures as provisional.
Overwhelmingly the sensor hardware. H04N pictorial-communication claims appear in 47 of the 59 records and H01L semiconductor-device claims in 19, while G06N AI-model computing claims appear in only 10. In practice this means the readout circuitry, pixel architecture and signal path carry most of the claimed novelty, with the learned model usually described as an application layer rather than the invention itself.
The corpus includes academic filers such as the University of Toronto and Dartmouth College, established sensor suppliers including Sony Semiconductor Solutions, and larger platform entrants such as Huawei, Google and ams-affiliated Heptagon Micro Optics. Notably, all of the named organisations tracked for recent-year momentum show zero filings in the latest tracked year, including a -100% year-on-year drop for Sony Semiconductor Solutions, which points to a lull in visible activity rather than sustained leadership by any single filer.
The gap between sensor-circuitry claims (47 of 59 records under H04N) and AI-model claims (only 10 under G06N) suggests under-claimed ground in architectures that tie the learned model directly into the pixel or readout path, rather than treating it as downstream processing. Specific thin areas include binary or quanta pixel readout for time-of-flight ranging, event-driven computational imaging, and AI-assisted holographic or computational microscopy, all of which appear in the corpus but with limited depth relative to core sensor-hardware claims.
Answers are grounded in the same dataset. Derived from a Patsnap search on CMOS Image Sensor AI and Machine Learning covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
Research CMOS Image Sensor AI and Machine Learning in depth with Eureka
Go past this page: query the whole cmos image sensor 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.