Real image denoising method based on multi-scale fusion and edge enhancement

Abstract

The invention discloses a real image denoising method based on multi-scale fusion and edge enhancement, and belongs to the technical field of computer vision images. In the image input stage, in orderto improve the generalization ability of the model, data enhancement is designed, and part of pixels of the content of an input noise map are randomly selected to be replaced with corresponding noiseless maps; multi-level smoothing is performed on the input noise image by using three convolution kernels with different receptive field sizes to obtain three preliminary smoothing results with different scales; a channel attention mechanism is used for carrying out adaptive expression on the multi-scale denoising result, and then fusion is carried out; an edge is extracted through a Laplace operator, the edge and texture information of an original noise image are introduced, detail enhancement of a fused smooth image is performed so that visual effect is improved. the invention is reasonablein design, maintains a higher operation speed while achieving a better denoising effect, and achieves a better effect on the denoising of a real image as a whole.

Description

technical field [0001] The invention relates to the technical field of computer vision images, in particular to a real image denoising method based on multi-scale fusion and edge enhancement. Background technique [0002] With the advancement of science and technology, mobile devices are becoming more and more popular, and image acquisition is becoming more and more convenient. Due to the use of relatively low-cost sensors and lenses, the images captured by mobile cameras such as mobile phone cameras are usually disturbed by noise, especially when the light is insufficient, the impact of noise is more serious, which will lead to a decrease in image quality. bring difficulty. Ensuring image quality is the basis for high-level vision applications such as object detection and semantic segmentation on images. Therefore, how to efficiently denoise real images and improve image quality is an important research topic in the field of computer vision. [0003] Real-Image Denoising...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to solve the problem that existing image denoising methods fail to pay attention to the variety and complexity of real noise, do not consider the different importance between feature channels, and fail to fully utilize multi-scale features. Therefore, the problem of relatively limited effects has been achieved, and a real image denoising method based on multi-scale fusion and edge enhancement is proposed that is reasonably designed, fully considers multi-scale information to improve the noise removal effect, and is relatively lightweight.

Images