A Joint Residual Learning and Structural Similarity Image Denoising Method

Abstract

The invention proposes an image denoising method combining residual learning and structure similarity. The present invention selects a plurality of high-definition images in the BSD database to construct a training data set and a test data set respectively; the training data set after cutting is obtained by centrally cutting the high-definition images in the training data set, and the training data set after cutting is preprocessed to obtain the preprocessing After training the data set, add a certain intensity of Gaussian noise to the preprocessed training data set and test data set, respectively, to obtain a noisy training data set and a noisy test data set; design a deep convolutional neural network, by designing the L2 norm and The loss function of the SSIM joint is minimized to train the deep convolutional neural network, and the clear image data set is obtained through the calculation between the noisy test data set and the noise residual image obtained according to the deep convolutional neural network. The advantage of the present invention is that the denoising effect is more in line with human visual perception.

Description

technical field [0001] The invention belongs to the fields of image processing and computer vision, and in particular relates to an image denoising method for joint residual learning and structural similarity. Background technique [0002] Image denoising has always been a research hotspot in the field of image processing. There are more and more ways to acquire images in practical applications, but during the acquisition and transmission of images, they will be affected by equipment and external factors, introducing various noises, making post-processing of images difficult, and noise information Seriously affects the human eye's understanding of image information. Therefore, it is particularly important to establish an image denoising method that conforms to human visual perception. [0003] The purpose of image denoising is to obtain a clear image after denoising from the noisy image to be processed. With the development of denoising algorithms, more obvious effects ha...

AI Technical Summary

Problems solved by technology

The model structure of this method is simple, and has achieved a certain denoising effect, but there are the following problems: (1) With the increase of network depth, it is difficult to directly learn the model of clear image after denoising; (2) the loss function only considers Although MSE can suppress large noise, it can tolerate small noise, and the effect after denoising is often inconsistent with human subjective

Images