Samsung · Filed Jun 26, 2026 · Published Nov 5, 2026

Samsung Files Patent for Restoring Lost Detail in Heavily Zoomed Phone Photos

Samsung's latest patent filing goes after the smooth, painted look that heavy digital zoom gives your photos, by adding the right kind of grain in the right places.

A distant photo of a building alongside an enlarged close-up highlights regions where fine textures blur and need enhancement. Drawing from patent filing US 2026/0350862 A1.
A distant photo of a building alongside an enlarged close-up highlights regions where fine textures blur and need enhancement.
See all 12 drawings from this filing ↓
Publication number US 2026/0350862 A1 US20260350862A1
Applicant SAMSUNG ELECTRONICS CO., LTD.
Filing date Jun 26, 2026
Publication date Nov 5, 2026
Inventors Yathish REDDY M, Sarvesh ., Gihun KIM, Yongjun KIM, Daehun KIM, Rishabh KAUSHIK, Addanki SANKARARAO, Raj Narayana GADDE
US classification 345/582
Status when we published Waiting for an examiner (Sep 1, 2026)
Parent application is a Continuation of PCTKR2026004831 (filed 2026-03-26)
Document 20 claims

What Samsung's texture patent does to zoomed photos

You're at a concert, zoomed all the way in on the singer, and the photo comes out looking like an oil painting. Skin is smooth, the backdrop is mush. That's what heavy digital zoom does: it crops and stretches the picture, and the fine texture gets lost.

Samsung's patent describes a way to put believable texture back. The phone first works out what's in the picture: skin, sky, clothing, a wall. Then it checks the lighting, how far you've zoomed, and what's moving. From that it decides how much grain each patch of the photo should get, and what kind.

A well-lit face or a bright window gets little or none, so it stays clean. A shadowy wall gets more, so it looks like a real wall. The goal is a zoomed photo that looks natural instead of fake.

From the filing · CLAIM 1
generating a texture map based on at least one of the visual frame or a segmentation map of the visual frame, wherein the texture map indicates a texture type for at least one region in the visual frame; …

Translation: The software categorizes the surfaces in your photo before trying to fix them.

How the phone decides where grain goes and how much

The system starts with a photo or a video frame, either live from the camera or saved on the phone. A segmentation step (software that labels every region of the picture, such as skin, sky, clothing or a window) sorts the scene. Using those labels and measurements of how rough each area is, it builds a texture map that records what kind of surface each region is.

Next it builds auxiliary maps, extra layers of context:

  • Illumination map: how bright each spot is, so bright areas can get less grain.
  • Zoom and weight map: uses the sensor's sensitivity (its gain) and the zoom level, so a dark shot that already has sensor grain doesn't get extra.
  • Optical flow map: shows which parts are moving, using the video frames and motion-sensor data.

All of these combine into a comprehensive weight map, a grain-strength setting for every pixel. The system then picks a noise sample for each region from a set of machine learning models, each tied to a texture type, such as film grain, camera-sensor noise or fractal noise (a smooth, natural-looking pattern). The chosen grain is blended into the frame.

In video, still areas like walls keep consistent grain from frame to frame to avoid flicker.

From the filing · THE ABSTRACT
… generating a comprehensive weight map based on the segmentation map, the texture map, and the at least one auxiliary map.

Translation: The phone combines all its analysis data to figure out how much digital noise to add.

Why Samsung cares about 100x zoom photos

If your phone camera could do this, the far end of the zoom would look less like a smeared painting. The filing points to phones without a dedicated telephoto lens, such as mid-range models, which depend on digital zoom for most of their reach. It also notes that flagship phones stretch telephoto shots to 30x or even 100x.

The filing stresses that the method is light enough to run on the phone itself, unlike heavier image generators it calls too demanding for real-time use. That matters for live video, where grain has to stay consistent from frame to frame. Keep in mind this is a patent application, and nothing in it says the idea will reach a product.

Samsung's 166th filing we've tracked in our camera sensor work since May builds on the thumbnail mismatch fix and the shadow and highlight capture fix.

Editorial take

Zoom is a headline camera spec, and the painted look at the far end is a complaint anyone who has zoomed in on a bird or a stage has run into. The filing itself talks about 30x and even 100x digital zoom, where the picture is mostly stretched guesswork. That's a real annoyance, though a cosmetic one: nobody loses anything important when a wall looks a bit smooth.

The fix matches the size of the problem. It adds believable grain in the right places, and the filing pitches it as light enough to run on the phone. It cannot bring back detail the lens never captured, so a readable license plate at 100x stays out of reach.

Grain tuned to skin, sky and shadows is a sensible polish step. The motion check that keeps grain from flickering in video is the part you'd most likely notice. Expect a modest improvement.

Get our take in your Top Stories

Liked this breakdown? Add Patentlyze as a preferred source on Google, and our plain-English take shows up more often in your Top Stories the next time Samsung patent news breaks.

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

12 drawing sheets from US 2026/0350862 A1 · click any drawing to enlarge

Blurred and smudged details caused by extreme digital zoom, drawing sheet 1 from US 2026/0350862 A1
Sheet 1 Blurred and smudged details caused by extreme digital zoom
Older methods add uniform grain across an entire picture, drawing sheet 2 from US 2026/0350862 A1
Sheet 2 Older methods add uniform grain across an entire picture
A smartphone sending image data to an external server, drawing sheet 3 from US 2026/0350862 A1
Sheet 3 A smartphone sending image data to an external server
Hardware components that process and enhance the photo, drawing sheet 4 from US 2026/0350862 A1
Sheet 4 Hardware components that process and enhance the photo
Different synthetic grain patterns matched to specific surfaces, drawing sheet 5 from US 2026/0350862 A1
Sheet 5 Different synthetic grain patterns matched to specific surfaces
Dividing a photo into distinct object regions, drawing sheet 6 from US 2026/0350862 A1
Sheet 6 Dividing a photo into distinct object regions
Combining identified regions to create custom surface details, drawing sheet 7 from US 2026/0350862 A1
Sheet 7 Combining identified regions to create custom surface details
Using motion sensors to chart lighting across every pixel, drawing sheet 8 from US 2026/0350862 A1
Sheet 8 Using motion sensors to chart lighting across every pixel
Show the other 4 sheets
Calculating how zoom level and sensor gain affect detail, drawing sheet 9 from US 2026/0350862 A1
Sheet 9 Calculating how zoom level and sensor gain affect detail
Tracking movement across the camera frame using motion sensors, drawing sheet 10 from US 2026/0350862 A1
Sheet 10 Tracking movement across the camera frame using motion sensors
Blending synthetic textures into the photo to restore lost detail, drawing sheet 11 from US 2026/0350862 A1
Sheet 11 Blending synthetic textures into the photo to restore lost detail
Steps for rebuilding lifelike textures in heavily magnified photos, drawing sheet 12 from US 2026/0350862 A1
Sheet 12 Steps for rebuilding lifelike textures in heavily magnified photos
Browse all the drawings 12 drawing sheets
Patent filing page

Source. Full patent text and figures from the official USPTO publication PDF.
Reader comments

Be the first to weigh in

Start the discussion

Real name or a handle, either is fine. Comments are read by a person before they appear, so allow a little time. Keep it about the filing.