# US 2026/0094249 A1 · Joint denoising and supersampling of graphics data

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

- Number: US 2026/0094249 A1
- Document ID: US20260094249A1
- Status: Published application
- Kind: A1
- Application number: 19/327,186
- Filed: 2025-09-12
- Published: 2026-04-02
- Earliest claimed: 2021-11-05
- Assignees: Intel Corporation
- Applicants: Intel Corporation
- Inventors: Manu Mathew Thomas; Karthik Vaidyanathan; Anton Kaplanyan; SungYe Kim; Gabor Liktor
- Web page: https://patentgrep.com/patent/US20260094249A1

## CPC

- G06T 5/70: Denoising; Smoothing
- G06F 18/213: Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06T 1/20: Processor architectures; Processor configuration, e.g. pipelining
- G06T 3/18: Image warping, e.g. rearranging pixels individually
- G06T 3/4046: using neural networks
- G06T 3/4053: based on super-resolution, i.e. the output image resolution being higher than the sensor resolution
- G06T 5/20: using local operators
- G06T 5/60: using machine learning, e.g. neural networks
- G06T 15/06: Ray-tracing
- G06T 2207/20081: Training; Learning
- G06T 2207/20084: Artificial neural networks [ANN]
- G06T 2210/36: Level of detail
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Joint denoising and supersampling of graphics data is described. An example of a graphics processor includes multiple processing resources, including a least a first processing resource including a pipeline to perform a supersampling operation; and the pipeline including circuitry to jointly perform denoising and supersampling of received ray tracing input data, the circuitry including first circuitry to receive input data associated with an input block for a neural network, second circuitry to perform operations associated with a feature extraction and kernel prediction network of the neural network, and third circuitry to perform operations associated with a filtering block of the neural network.

## Claims

### Claim at position 1

1-20. (canceled)

### Claim 21

A graphics processor comprising:
a plurality of processing resources, including a least a first processing resource including a pipeline to perform a supersampling operation; and
the pipeline including circuitry to jointly perform denoising and supersampling of received ray tracing input data in a neural network, the neural network including:
a feature extraction network to receive ray tracing input data from an input block, the feature extraction network to process the received ray tracing input data to extract feature data from the ray tracing input data, the processing including both denoising and supersampling; and
a filtering block to receive the ray tracing input data and the feature data for filtering operation, the filtering block including a plurality of filtering paths for filtering of data, each filter path filtering a component of the ray tracing data.

### Claim 22 (depends on 21)

The graphics processor of claim 21, wherein the plurality of filtering paths includes:
an albedo filtering path;
a specular filtering path; and
a diffuse filtering path.

### Claim 23 (depends on 22)

The graphics processor of claim 22, wherein the albedo filtering path includes albedo and normal filtering.

### Claim 24 (depends on 21)

The graphics processor of claim 21, wherein an output from the plurality of filtering paths is fed back from the filtering block to the input block.

### Claim 25 (depends on 24)

The graphics processor of claim 24, wherein the filtering block further receives previous output signals from the input block for filtering operation.

### Claim 26 (depends on 21)

The graphics processor of claim 21, wherein:
the feature extraction network includes a plurality of convolution layers; and
the plurality of convolution layers operate at a first precision and the filtering block operates at a second precision, the first precision being lower than the second precision.

### Claim 27 (depends on 26)

The graphics processor of claim 26, wherein the first precision includes either INT4 or INT8.

### Claim 28

A method comprising:
receiving ray tracing input data at a processing resource, the processing resource including a pipeline to perform a supersampling operation;
performing feature extraction processing of the received ray tracing input data to generate feature data, the processing including both denoising and supersampling of the received ray tracing input data;
receiving the ray tracing input data and the feature data for filtering operation;
performing the filtering operation in a plurality of filtering paths to generate filtered data, each filter path filtering a component of the ray tracing data; and
generating output signals based at least in part on the filtered data.

### Claim 29 (depends on 28)

The method of claim 28, wherein the plurality of filtering paths includes:
an albedo filtering path;
a specular filtering path; and
a diffuse filtering path.

### Claim 30 (depends on 29)

The method of claim 29, wherein the albedo filtering path includes albedo and normal filtering.

### Claim 31 (depends on 28)

The method of claim 28, further comprising:
feeding back the output signals to an input block.

### Claim 32 (depends on 31)

The method of claim 31, further comprising:
further receiving previous output signals for filtering operation.

### Claim 33 (depends on 28)

The method of claim 28, wherein:
the feature extraction operation includes operation of a plurality of convolution layers; and
the plurality of convolution layers operate at a first precision and the filtering operation operates at a second precision, the first precision being lower than the second precision.

### Claim 34 (depends on 33)

The method of claim 33, wherein the first precision includes either INT4 or INT8.

### Claim 35

One or more non-transitory computer-readable storage mediums having stored thereon executable computer program instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
receiving ray tracing input data at a processing resource, the processing resource including a pipeline to perform a supersampling operation;
performing feature extraction processing of the received ray tracing input data to generate feature data, the processing including both denoising and supersampling of the received ray tracing input data;
receiving the ray tracing input data and the feature data for filtering operation;
performing the filtering operation in a plurality of filtering paths to generate filtered data, each filter path filtering a component of the ray tracing data; and
generating output signals based at least in part on the filtered data.

### Claim 36 (depends on 35)

The one or more non-transitory computer-readable storage mediums of claim 35, wherein the plurality of filtering paths includes:
an albedo filtering path;
a specular filtering path; and
a diffuse filtering path.

### Claim 37 (depends on 36)

The one or more non-transitory computer-readable storage mediums of claim 36, wherein the albedo filtering path includes albedo and normal filtering.

### Claim 38 (depends on 35)

The one or more non-transitory computer-readable storage mediums of claim 35, further comprising instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
feeding back the output signals to an input block.

### Claim 39 (depends on 38)

The one or more non-transitory computer-readable storage mediums of claim 38, further comprising instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
further receiving previous output signals for filtering operation.

### Claim 40 (depends on 35)

The one or more non-transitory computer-readable storage mediums of claim 35, wherein:
the feature extraction operation includes operation of a plurality of convolution layers; and
the plurality of convolution layers operate at a first precision and the filtering operation operates at a second precision, the first precision being lower than the second precision.
