Image denoising model training method, image denoising method and device and medium

Abstract

The invention relates to an image denoising model training method, an image denoising method and device and a medium, and the training method comprises the steps: collecting a plurality of sample image groups through photographing equipment, obtaining the sensitivity of each sample image group, and determining a noise representation image corresponding to each sample image group according to the sensitivity; determining a training input image group and a target image associated with each sample image group, wherein each training input image group comprises all or part of sample images in the corresponding sample image group and corresponding noise characterization images; constructing a plurality of training pairs, wherein each training pair comprises a training input image group and a corresponding target image; and training the image denoising model by using the plurality of training pairs until the image denoising model converges. According to the invention, the sensitivity information of the image is used as a part of the training input data of the image denoising model, so that the image denoising model learns the noise information of more dimensions, and the image noise can be effectively removed after the image denoising model is successfully trained.

Description

technical field [0001] The present disclosure relates to mobile terminal data processing technology, and in particular to a training method of an image denoising model, an image denoising method, a device and a medium. Background technique [0002] Mobile terminals generally have a camera function. The insufficient sampling rate of the image sensor in the mobile terminal, and the noise from various aspects such as image acquisition, transmission, and compression lead to the loss of details and excessive color noise in the image captured by the mobile terminal. This problem is also present in images and videos captured in low-light environments, and is exacerbated in images captured by smartphones with small-aperture cameras. In addition, in the process of image acquisition, aliasing phenomenon will be caused when the sampling rate is low. [0003] With the continuous improvement of users' demands on image effects, how to further reduce image noise is a technical problem th...

AI Technical Summary

Problems solved by technology

The sampling rate of the image sensor in the mobile terminal is insufficient, and the noise from various aspects such as image acquisition, transmission, and compression leads to the loss of details and excessive color noise in the image captured by the mobile terminal.

This issue is also present in images and videos captured in low light environments, and is more severe in images captured by smartphones with small aperture cameras

In addition, in the process of image acquisition, when the sampling rate is low, it will also cause aliasing.

Images