Joint denoising and demosaicking method for color RAW images guided by monochrome images

A joint denoising and demosaicking method for a color RAW image guided by a monochrome image, wherein the method performs denoising and demosaicking on color RAW images under a guidance of the monochrome image in parallax scenarios, comprising:

• step S1, dataset construction:
• constructing a synthetic noisy dataset of a monochrome-color binocular camera system in a dark scene;
• step S2, constructing a joint denoising and demosaicking model based on the dataset in the step S1, wherein the joint denoising and demosaicking model comprises an aligned guidance image generation module and a guided denoising and demosaicking module;
• first, constructing the aligned guidance image generation module, computing, by the aligned guidance image generation module, a feature correlation between the monochrome image and the RAW image in a parallax direction based on a parallax attention mechanism, and constructing an aligned guidance image feature by aggregating a relevant structural feature to obtain an aligned guidance image;
• and using a clean grayscale image ground truth corresponding to the RAW image as supervision, and training the aligned guidance image generation module until convergence based on a perceptual loss function;
• and
• next, constructing the guided denoising and demosaicking module, extracting, by the guided denoising and demosaicking module, a feature of the RAW image and a feature of the aligned guidance image, respectively, performing feature fusion through channel concatenation among a feature dimension, generating a clean RGB image corresponding to the RAW image through a feature decoder, and completing a guided denoising process and a guided demosaicking process simultaneously;
• and training the guided denoising and demosaicking module until convergence based on a color-structure joint loss function;
• and
• step S3, denoising and demosaicking the color RAW image based on the joint denoising and demosaicking model constructed in the step S2.