Method and apparatus for denoising a low-light image

An apparatus for denoising a low-light image, wherein the low-light image contains noise and dead pixels, comprising:

• an acquisition module configured to acquire a low-light image in RAW domain;
• a denoising module configured to perform a preset image enhancement transformation on the low-light image in the RAW domain, and input the transformed image into a pre-trained denoising network model to obtain an output image;
• wherein, the denoising network model is trained based on sample images, and the sample images comprise a low-light image with simulated dead pixels and a noiseless image;
• an inverse transformation module configured to perform an inverse transformation of the preset image enhancement transformation on the output image to obtain a denoised image;
• wherein the denoising module comprises an enhancement transformation sub-module, and the enhancement transformation sub-module is specifically configured to:
• sequentially perform normalization and gamma transformation on the low-light image in the RAW domain;
• wherein the normalization is performed by a formula:
• x
• ij
• ′
• =
• x
• ij
• max
• ⁢
• _b
• wherein xij denotes a pixel value of a pixel with coordinates (i, j) in the low-light image in the RAW domain, max_b denotes a maximum brightness of pixels in the low-light image in the RAW domain, x′ij denotes a pixel value of the pixel with coordinates (i, j) in the normalized low-light image in the RAW domain;
• wherein the inverse transformation sequentially includes inverse gamma transformation and inverse normalization transformation;
• wherein the noiseless image is obtained by:
• adjusting an ISO value of an imaging device to lowest for shooting, so as to obtain a noiseless RAW image:
• dividing pixel values of pixels in the noiseless RAW image by different ratios, to obtain low-light noiseless images with different levels, which are determined as noiseless images in the sample images.

An electronic device comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus;

• the memory is configured to store computer programs;
• the processor is configured to execute the programs stored in the memory to perform the following operations:
• acquiring a low-light image in a RAW domain, wherein the low-light image contains noise and dead pixels;
• performing a preset image enhancement transformation on the low-light image in the RAW domain, and inputting the transformed image into a pre-trained denoising network model to obtain an output image;
• wherein, the denoising network model is trained based on sample images, and the sample images comprise a low-light image with simulated dead pixels and a noiseless image;
• performing an inverse transformation of the preset image enhancement transformation on the output image to obtain a denoised image;
• wherein performing the preset image enhancement transformation on the low-light image in the RAW domain comprises:
• sequentially performing normalization and gamma transformation on the low-light image in the RAW domain,
• wherein the normalization is performed by a formula:
• x
• ij
• ′
• =
• x
• ij
• max
• ⁢
• _b
• wherein xij denotes a pixel value of a pixel with coordinates (i, j) in the low-light image in the RAW domain, max_b denotes a maximum brightness of pixels in the low-light image in the RAW domain, x′ij denotes a pixel value of the pixel with coordinates (i, j) in the normalized low-light image in the RAW domain;
• wherein the inverse transformation sequentially includes inverse gamma transformation and inverse normalization transformation;
• wherein the noiseless image is obtained by:
• adjusting an ISO value of an imaging device to lowest for shooting, so as to obtain a noiseless RAW image:
• dividing pixel values of pixels in the noiseless RAW image by different ratios, to obtain low-light noiseless images with different levels, which are determined as noiseless images in the sample images.