Google Patents a Tap-to-Cut Photo System That Uses Distance Data to Isolate Objects
Google is filing a patent for a photo system that lets you tap any object in a picture and get a clean cutout automatically, using depth data your phone's camera already captures to do the heavy lifting.
What Google's depth-map object cutout actually does
Every time you snap a portrait on a modern phone, the camera records more than color: it also logs how far away each part of the scene is. That distance data, called a depth map, normally sits in the background doing things like blurring the background in portrait mode.
Google's patent puts that depth data to a new use. When you tap on an object in a photo, the system reads the depth map to figure out which pixels belong to that object and which belong to everything else. It then draws a precise selection around just the object you tapped, and hands you a clean cutout you can pull out and use however you like.
The result is the kind of one-tap object selection that currently requires either careful manual tracing or a powerful AI model running in the cloud. Here, your phone's own depth sensor does most of the work, locally, on the image file itself.
… receiving user input indicative of selection of an object depicted in the primary image; generating a segmentation mask based on the depth map, wherein the segmentation mask identifies a subset of pixels of the primary image that correspond to the object …
Translation: The system uses your tap to figure out which pixels belong to the object you want to cut out.
How the depth map turns a tap into a pixel-perfect selection
The patent describes a full pipeline, from capture to edit. When a photo is taken, the camera records a primary image alongside depth values for each pixel. Those depth values are stored as a depth map inside the image file's metadata, packaged in a structured format that includes a profile element (a tag that tells software what kind of image this is) and separate camera elements.
When you open the photo later and tap an object, the system reads that stored depth map and generates a segmentation mask (essentially a stencil that marks which pixels belong to your chosen object and which don't). Because depth information already separates foreground from background in three-dimensional space, the system can draw that stencil without having to analyze colors or edges the way a traditional selection tool would.
The key steps the claim covers are:
- Loading a photo file that contains both pixel data and a depth map in its metadata
- Displaying the photo and accepting a tap on an object
- Using the depth map to generate a segmentation mask around that object
- Extracting the object as a separate image
- Presenting that cutout image in a user interface for further use
The patent also covers the encoding side: a defined file structure that stores the depth map alongside the primary image so it's available for exactly this kind of editing later.
… image metadata that includes a device element that includes a profile element indicative of an image type and a first camera element, wherein the first camera element includes an image element and a depth map based on the depth values …
Translation: The file stores both the picture and its distance data inside a standardized container format.
What this means for editing photos on your phone
For everyday users, this could make object cutouts as easy as tapping a screen. Right now, cleanly isolating a subject in a photo either takes manual effort in an app like Photoshop, or relies on cloud-based AI tools that send your image to a server. A system that does the job from depth data already stored in the file would be faster and keep everything on-device.
The patent also covers the storage format, not just the editing step. That means the pattern in Google's computational photography filings points toward a system where depth data is a first-class citizen in the image file from the moment you press the shutter, ready to power editing features later rather than being discarded or siloed in a separate app.
This is the 53rd Google patent we've tracked since May in the AI photo editing race, building on earlier filings like one on blending two frames for video prediction and one on using color data to fix blur.
The patent covers any method that captures a photo, packages depth information alongside it in the same file, and then displays that photo. Nothing in the claim limits it to a specific camera type, depth-sensing approach, or screen. That scope is remarkably wide.
The one hard boundary the claim draws is about timing and storage: the depth data must be encoded inside the file at the moment of saving, not added later or kept in a separate file. Any photo pipeline that discards depth before saving, or stores it elsewhere, falls outside the claim.
That boundary matters because depth baked into a photo file is what powers reliable background blur, subject cutouts, and similar editing tools. Owning the step where depth gets locked into the image means having a hand in how cameras, photo apps, and editing software are built from the ground up.
There are more where this came from
We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.
The drawings
6 drawing sheets from US 2026/0281298 A1 · click any drawing to enlarge
Want this weekly breakdown for a company we don't cover? Patentlyze Pro →