# US 2023/0058096 A1 · Method and system for denoising using neural networks > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2023/0058096 A1 - Document ID: US20230058096A1 - Status: Published application - Kind: A1 - Application number: 17/678,438 - Filed: 2022-02-23 - Published: 2023-02-23 - Earliest claimed: 2021-08-19 - Assignees: GoPro, Inc. - Applicants: GoPro, Inc. - Inventors: Matias Tassano Ferrés; Nicolas Rahmouni; Eva Coupeté; Thomas Armel Derbanne - Web page: https://patentgrep.com/patent/US20230058096A1 ## CPC - G06T 5/70: Denoising; Smoothing - G06T 5/60: using machine learning, e.g. neural networks - G06T 7/40: Analysis of texture (depth or shape recovery from texture G06T7/529) - G06T 2207/10024: Color image - G06T 2207/20081: Training; Learning - G06T 2207/20084: Artificial neural networks [ANN] - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL ## Abstract Methods and systems for denoising. A method for denoising includes receiving an image from an image sensor, denoising the image, in a non-linear domain by a denoiser, by applying a noise map to the image to obtain a denoised image. Training losses for the denoiser are processed in a linear domain. The method includes storing, displaying, or transmitting an output image based on the denoised image. ## Claims ### Claim 1 A method for denoising, the method comprising: receiving an image from an image sensor; denoising the image, in a non-linear domain by a denoiser, by applying a noise map to the image to obtain a denoised image, wherein training losses for the denoiser are processed in a linear domain; and storing, displaying, or transmitting an output image based on the denoised image. ### Claim 2 (depends on 1) The method of claim 1, wherein the denoising the image further comprises: applying a re-noising factor to the image to obtain the denoised image. ### Claim 3 (depends on 1) The method of claim 1, wherein texture map processing is accounted for during denoiser training. ### Claim 4 (depends on 1) The method of claim 1, further comprising: processing, in the non-linear domain by the denoiser in an offline configuration, a noisy image from a training dataset to obtain a training denoised image; comparing, in a linear domain, the training denoised image with a reference image; and optimizing the denoiser based on the comparison of the training denoised image with the reference image. ### Claim 5 (depends on 4) The method of claim 4, wherein the noise map is factored with a unity value. ### Claim 6 (depends on 4) The method of claim 4, further comprising: determining a loss between the training denoised image and the reference image, wherein the loss accounts for a texture map. ### Claim 7 (depends on 4) The method of claim 4, further comprising: setting a factor for a texture map which provides a greater weight to one of flat or texture in the noisy image; and determining a texture map-based loss between the training denoised image and the reference image. ### Claim 8 A method for denoising, the method comprising: denoising an image, in a first color space by a denoiser, by applying a re-noising factor to the image to obtain a denoised image, wherein training losses for the denoiser are processed in a second color space; and storing, displaying, or transmitting an output image based on the denoised image. ### Claim 9 (depends on 8) The method of claim 8, wherein the denoising the image further comprising: applying a noise map to the image to obtain the denoised image. ### Claim 10 (depends on 8) The method of claim 8, wherein texture map weighting is accounted for during denoiser training. ### Claim 11 (depends on 8) The method of claim 8, further comprising: processing, in the first color space by the denoiser in a training configuration, a noisy image from a training dataset; comparing, in a second color space, a training denoised image with a reference image; and optimizing the denoiser based on the comparison of the training denoised image with the reference image, wherein the re-noising factor is disabled during denoiser training. ### Claim 12 (depends on 11) The method of claim 11, wherein a noise map applied during denoising is factored with a unity value during training. ### Claim 13 (depends on 11) The method of claim 11, further comprising: determining a loss between the training denoised image and the reference image, wherein the loss accounts for a texture map weighting. ### Claim 14 (depends on 11) The method of claim 11, further comprising: setting a factor for a texture map which provides a greater weight to one of flat or texture in the noisy image; and determining a texture map-based loss between the training denoised image and the reference image. ### Claim 15 (depends on 11) The method of claim 11, wherein the first color space is YUV. ### Claim 16 (depends on 15) The method of claim 15, wherein the second color space is RGB. ### Claim 17 An image capture device, comprising: an image sensor configured to detect an image; and an image processor configured to receive the image in a first color domain and comprising a denoiser configured to denoise the image to obtain a denoised image in the first color domain, wherein weights and training losses for the denoiser are processed in a second color domain during an offline configuration and the weights are saved to the image capture device, and wherein the image processor is configured to store, display, or transmit an output image based on the denoised image. ### Claim 18 (depends on 17) The image capture device of claim 17, wherein the denoiser is further configured to: apply a re-noising factor to the image to obtain the denoised image. ### Claim 19 (depends on 18) The image capture device of claim 18, wherein the denoiser is further configured to: apply a noise map to the image to obtain the denoised image. ### Claim 20 (depends on 19) The image capture device of claim 19, wherein the training losses account for a texture map weighting selection which emphasizes one of flat or texture in the image.