# US 2026/0073474 A1 · Predictive color image processing

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0073474 A1
- Document ID: US20260073474A1
- Status: Published application
- Kind: A1
- Application number: 19/322,255
- Filed: 2025-09-08
- Published: 2026-03-12
- Earliest claimed: 2021-08-31
- Assignees: NVIDIA Corporation
- Applicants: NVIDIA Corporation
- Inventors: Gopal Triplicane Venkatesan
- Web page: https://patentgrep.com/patent/US20260073474A1

## CPC

- G06T 3/4015: Image demosaicing, e.g. colour filter arrays [CFA] or Bayer patterns
- G06T 3/4046: using neural networks
- G06T 5/20: using local operators
- G06T 2207/10024: Color image
- G06T 2207/20081: Training; Learning
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL

## Abstract

Apparatuses, systems, and techniques to process image data. In at least one embodiment, a neural network is trained to perform demosacing of two-dimensional image data obtained from an image sensor.

## Claims

### Claim at position 1

-40. (canceled)

### Claim 41

A processor, comprising one or more circuits to generate a final image based, at least in part, on sampling different colors from an image according to different frequencies, wherein the final image comprises one or more missing color values added to the image by one or more neural networks.

### Claim 42 (depends on 41)

The processor of claim 41, wherein the one or more missing color values correspond to different colors of a Bayer filter.

### Claim 43 (depends on 42)

The processor of claim 42, wherein a green color of the different colors of the Bayer filter is sampled with a first one of the different frequencies and wherein a red color and blue color with a second one of the different frequencies.

### Claim 44 (depends on 41)

The processor of claim 41, wherein the one or more missing colors are determined according to output from the one or more neural networks input to a notch filter that filters the input at the different frequencies.

### Claim 45 (depends on 44)

The processor of claim 44, wherein the notch filter is a two dimensional filter.

### Claim 46 (depends on 41)

The processor of claim 41, wherein the image is a raw image captured using a sensor with a Bayer filter.

### Claim 47 (depends on 41)

The processor of claim 41, wherein the one or more color values correspond to a quad in the image and wherein the one or more neural networks use data in a region around the quad to compute the one or more missing color value values.

### Claim 48

A method, comprising:
generating, by one or more processors, a final image based, at least in part, on sampling different colors from an image according to different frequencies, wherein the final image comprises one or more missing color values added to the image by one or more neural networks.

### Claim 49 (depends on 48)

The method of claim 48, wherein the one or more missing color values correspond to different colors of a Bayer filter.

### Claim 50 (depends on 49)

The method of claim 49, wherein a green color of the different colors of the Bayer filter is sampled with a first one of the different frequencies and wherein a red color and blue color with a second one of the different frequencies.

### Claim 51 (depends on 48)

The method of claim 48, wherein the one or more missing colors are determined according to output from the one or more neural networks input to a notch filter that filters the input at the different frequencies.

### Claim 52 (depends on 51)

The method of claim 51, wherein the notch filter is a two dimensional filter.

### Claim 53 (depends on 48)

The method of claim 48, wherein the image is a raw image captured using a sensor with a Bayer filter.

### Claim 54 (depends on 48)

The method of claim 48, wherein the one or more color values correspond to a quad in the image and wherein the one or more neural networks use data in a region around the quad to compute the one or more missing color value values.

### Claim 55

A system, comprising one or more processors to generate a final image based, at least in part, on sampling different colors from an image according to different frequencies, wherein the final image comprises one or more missing color values added to the image by one or more neural networks.

### Claim 56 (depends on 55)

The system of claim 55, wherein the one or more missing color values correspond to different colors of a Bayer filter.

### Claim 57 (depends on 56)

The system of claim 56, wherein a green color of the different colors of the Bayer filter is sampled with a first one of the different frequencies and wherein a red color and blue color with a second one of the different frequencies.

### Claim 58 (depends on 55)

The system of claim 55, wherein the one or more missing colors are determined according to output from the one or more neural networks input to a notch filter that filters the input at the different frequencies.

### Claim 59 (depends on 55)

The system of claim 55, wherein the image is a raw image captured using a sensor with a Bayer filter.

### Claim 60 (depends on 55)

The system of claim 55, wherein the one or more color values correspond to a quad in the image and wherein the one or more neural networks use data in a region around the quad to compute the one or more missing color value values.
