Computer Vision Patents: Top Companies & Filing Trends 2026
- Filing is not consolidated. the ranked leader holds 4,065 records but the top 10 combined account for only 18.8% of all 160,812 records in scope, leaving a long tail of single- and few-filing entrants.
- Momentum is cooling among the largest filers. recent-year filings for leading assignees are down sharply year-on-year, though the newest years are still filling in as publications catch up with filing dates.
- Classification spreads across recognition and processing, not one bucket. G06T (image data processing) leads at 4.3% of records, followed closely by pictorial communication and image/video recognition classes, none dominating outright.
Filing growth compares 2021 (2,204 records) with 2024 (1,622) — 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 160,812 records in scope (CR5), not by the ranked leaders only.
What the computer vision patent record shows
The search set covers 160,812 published records filed or published between 2015 and the 2026 data cut-off, spanning image sensing, feature detection, object recognition, image reconstruction, loss-function design and model architecture claims. Filing activity rose through the late 2010s, peaked in 2022 at 2,257 records, and has since pulled back — filings fell from 2,204 in 2021 to 1,622 in 2024, a decline of 26% over that three-year span, the most recent period the data can treat as complete.
Because publication trails filing by roughly 18 months, the 2025 and 2026 counts in the trend chart are undercounts, not evidence of a further slowdown. The assignee ranking spans 100 companies, none of which controls the field outright: the top 5 combined hold 11.2% of all records, and the top 10 combined hold 18.8%. That leaves the majority of the corpus distributed across a long tail of filers, which is where much of the unclaimed technical space sits.
Let an AI agent run this analysis on your own technology
Pick a task. Every answer cites the patents behind it.
Filing trends and technology composition
The filing curve and the IPC breakdown together show a field that grew fast, peaked, and now spreads its claims across several adjacent subclasses rather than one dominant category.
A decade of filing activity, with a 2022 peak
Annual filings rose from 1,032 in 2017 to a peak of 2,257 in 2022, then eased to 1,622 by 2024 — a 26% drop from the 2021 level. Treat 2025 and 2026 figures as provisional; publication lag means recent filings are still arriving in the record.
Recognition and processing classes lead, but none dominates
G06T (image data processing and generation) tops the IPC breakdown at 4.3% of the 160,812 records in scope, with H04N (pictorial communication), G06V (image/video recognition) and G06K (data recognition) each close behind. G06N (AI-based computing) and G06F (digital data processing) sit lower but still material, while G02B (optical elements) and G06Q (commerce applications) mark the edges of the core claim space.
Shares are the percentage of the 160,812 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Go deeper on Computer Vision Patent Landscape with Eureka
This page is one run against one query. Ask Eureka your own question about computer vision patent landscape and every answer comes back with the patent numbers behind it.
Try EurekaThe most-cited records anchoring this field
Selection of object recognition models for computer vision
This filing, assigned to Salesforce, describes a compliance-auditing system that receives an audit image from a mobile device and selects an object recognition model from a model selection list based on a required object recognition list, matching a model identifier to the objects a given audit needs to recognise. The approach frames object recognition as a model-selection problem rather than a single fixed-model pipeline, which is the detail worth checking against any multi-model recognition architecture.Filed 2021-08-05, published as US20210240967A1.
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US20160026253A1 | Methods and systems for creating virtual and augmented reality | 3,434 |
| 2 | US20140201126A1 | Methods and Systems for Applications for Z-numbers | 1,970 |
| 3 | US20040105264A1 | Multiple Light-Source Illuminating System | 1,261 |
| 4 | US20090143141A1 | Intelligent Multiplayer Gaming System With Multi-Touch Display | 1,150 |
| 5 | US20110161076A1 | Intuitive Computing Methods and Systems | 1,137 |
| 6 | US20100046842A1 | Methods and Systems for Content Processing | 1,088 |
| 7 | US20110212717A1 | Methods and Systems for Content Processing | 1,081 |
| 8 | US20190258251A1 | Systems and methods for safe and reliable autonomous vehicles | 1,044 |
| 9 | US20190339688A1 | Methods and systems for data collection, learning, and streaming of machine signals for analytics and mainten… | 974 |
| 10 | US20100111370A1 | Method and apparatus for estimating body shape | 957 |
Citation counts inside a searched corpus favour older, foundational filings — read them as a signal of influence on later claim drafting, not as a ranking of current commercial importance.
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 →Reading the concentration and momentum data
Three patterns matter for anyone deciding where to file next or who to watch: how concentrated ownership is, how fast the leaders are still filing, and where the technology classes overlap.
Ownership is spread thin at the top
The leading assignee holds 4,065 records, and the fifth-ranked holds 3,207, but the combined top 5 still account for only 11.2% of the full record set. Filing rights in computer vision are not concentrated the way they are in some hardware-heavy fields.
The 2022 peak has not been sustained
Filings dropped from 2,204 in 2021 to 1,622 in 2024. Several of the largest current filers also show sharp year-on-year declines in the most recent tracked year, though that year is still incomplete in the published record.
Claims cluster around image processing and recognition
G06T, H04N, G06V and G06K together capture the bulk of classified activity, each in the 2.9%-4.3% range of the 160,812 records in scope. No single subclass dominates, which suggests claim drafting still has room to differentiate within adjacent categories.
Filing is anchored in the US, with Europe and PCT next
The United States receiving office accounts for the largest single share of filings at 13,459, ahead of the EPO at 2,784 and WIPO/PCT at 2,485. UK, Australian and Canadian receiving offices trail well behind.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to computer vision patent landscape, with the prior art for and against each one.
Who is filing, and where the gaps sit
The ranked leaders include large diversified technology companies alongside device and platform specialists, but the ranking's long tail is where much of the field's white space is concentrated.
The top-ranked filer sets the pace but not the ceiling
The leading assignee's 4,065 records represent a meaningful lead over fifth place at 3,207, yet the gap narrows quickly further down the ranking, and the leader's share of the full 160,812-record corpus remains modest in absolute terms.
Recent-year filings are down across the board for large filers
Every large assignee tracked for recent-year momentum shows a steep year-on-year decline in the latest tracked year, from roughly -73% to -98%. This is consistent with publication lag rather than a confirmed pullback, since the most recent year is still being filled in.
Collaboration is limited and concentrated in a few pairs
Only 10 co-assignee pairs appear in the data, with the strongest pairings linking a large semiconductor filer to individual named inventors rather than to other corporate assignees. Cross-company joint filing is not a defining feature of this field.
| Assignee | Recent year | YoY |
|---|---|---|
| NVIDIA Corp | 12 | -81% |
| Google LLC | 4 | -73% |
| Snap Inc | 4 | -83% |
| Huawei Technologies Co., Ltd. | 2 | -85% |
| Qualcomm Inc | 1 | -98% |
| Samsung Electronics Co., Ltd. | 1 | -88% |
| Microsoft Technology Licensing, LLC | 1 | -90% |
| Semiconductor Components Industries, LLC | 0 | — |
Where to take this analysis
The dataset points to specific next steps depending on whether the goal is freedom-to-operate, competitive tracking, or identifying open claim space.
Check freedom-to-operate against the most-cited records
The highest-citation filings in this corpus, including the augmented and virtual reality record cited over 3,400 times, anchor a large share of downstream claim language. Any new filing in overlapping subclasses should be checked against these first.
Explore citation trees in EurekaTrack the leaders' slowing filing rate over the next publication cycle
Several top assignees show sharp year-on-year declines in the latest tracked year, but that year is not yet complete in the published record. Re-checking this trend once 2025-2026 filings finish publishing will show whether the pullback is real or an artefact of lag.
Set up assignee monitoring in EurekaScope a first filing in an under-claimed sub-area
Loss-function design, sensor fusion for edge devices, and multi-model selection frameworks all show thinner ownership concentration than the core recognition classes. A claim scoped around one of these branches faces less dense prior art.
Draft claim scope in EurekaCommon questions about the computer vision patent landscape
The ranking of 100 assignees in this dataset shows the leader holding 4,065 records, with the next several places trailing but still substantial, down to 3,207 at fifth place and 2,100 at tenth. However, the top 10 combined account for only 18.8% of the full 160,812-record corpus, so no single company or small group controls the field. The broader ranking includes a long tail of companies with far fewer filings each, which is typical for a fast-growing, technically diverse field like this one.
Filing activity rose steadily through the late 2010s and peaked in 2022 at 2,257 records, then declined to 1,622 by 2024, a drop of 26% from the 2021 level. That is the most recent period the data can be treated as complete, since publication typically lags filing by around 18 months. The 2025 and 2026 figures in any trend chart should be read as provisional undercounts rather than confirmation of a continued slowdown.
The dataset's IPC breakdown shows G06T (image data processing and generation) as the largest single class at 4.3% of the 160,812 records in scope, followed closely by H04N (pictorial communication), G06V (image and video recognition) and G06K (data recognition and presentation), each in the 2.9%-3.6% range. G06N (AI-based computing) and G06F (digital data processing) are also material but smaller shares. Because a single record can carry multiple IPC classes, these shares add up to more than 100% and should not be summed into a single total.
The clearest signal of open claim space is the gap between the dense core classes (image processing, recognition, pictorial communication) and adjacent branches with thinner ownership, such as loss-function design for image reconstruction, sensor-fusion feature detection for edge devices, and multi-model selection frameworks for recognition pipelines. These sit within the same search scope but show less concentrated filing in the ranked assignee data. A first filing scoped tightly to one of these branches is likely to face less crowded prior art than a filing aimed squarely at core object-recognition claims.
Citation counts accumulate over time inside a searched corpus, so older filings such as the augmented and virtual reality record published in 2016 and cited more than 3,400 times naturally outrank recent filings that have not had time to accumulate citations. This is a signal of historical influence on how later claims were drafted, not a measure of which technology is most commercially important today. Newer filings addressing the same technical problem may be equally or more relevant to a current freedom-to-operate check even with a lower citation count.
Research Computer Vision Patent Landscape in depth with Eureka
Go past this page: query the whole computer vision patent landscape 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.