Application of Image Denoising Method Based on Two-Way Coupling Diffusion Equation in Public Security Forensics

Abstract

This paper uses the web live broadcast and on-demand platform based on the B/S architecture as the application side and designs a video image forensic system that can meet multiple police types and multiple application scenarios. The system uses mobile phones as the video image capture terminal to solve the problem of rapid response and concealment and uses 5G communication technology as the transmission medium to solve the problem of device mobility and link maintenance. The problem of diversification of the use and application modes of multiple police types is solved; the video image evidence is managed in a centralized storage, audit, and export method, and the security and authenticity of the evidence are solved. While the system realizes a series of functions such as the collection, transmission, storage, and application of video image evidence, it also realizes the application-side video image live broadcast function according to actual work needs and solves the large-scale case command and decision-making problem that has been plagued by public security organs. In order to remove the noise in the public security forensic images and to smooth the noise while retaining the details of the image, this paper proposes a denoising algorithm based on the two-way coupling diffusion equation. By improving the second-order partial differential equation, a new diffusion function with better diffusion effect than the original model is constructed. We combined the adaptive edge threshold and stop criterion to establish a new denoising algorithm model, which can get better denoising results. When the noise level is low, the PSNR value and SSIM value of several denoising methods are relatively ideal, and the result is at a higher level, the denoising picture effect is better, and there is no obvious incomplete noise removal or detail problems. As the noise level increases, the denoising results will gradually decrease, and the effects will also vary to different degrees. When the noise intensity increases, visually, it can be clearly seen that the two-way coupled diffusion equation and DnCNN have better denoising effects. When the noise level is high, the two-way coupled diffusion equation network is used to use the clear image and the denoised image for indistinguishable calculation. The method in this paper almost retains all the texture details in the clear image, and there are almost no artifacts and images. On the other hand, the color of the image after denoising by the method in this paper is more vivid, and it is closer to the target picture in terms of picture definition and tone, the denoising effect is ideal, and the generated image has a higher degree of restoration. Compared with the residual GAN, the two-way coupling diffusion equation network converges faster and the network performance is improved.