The Security Image Denoising Algorithm Based On Partial Differential Equation

Security image is a typical kind of X-ray image. As a result of of X-ray imaging principle and the complexity of luggage bags, security image contains many details and noises, which troubles the security staff so much. The detail of security image is so important, which is relate to the judgment of security staff. However the traditional denoising algorithms cannot reduce the noise effectively and protect the detail when they deal with security image.To solve the problem, this thesis researched a kind of adaptive image denoising algorithm based on partial differential equation. It will be used to denoise the security image which is influenced by the noise. There are different kinds of PDE algorithms, this thesis emphasizes the Perona-Malik (P-M) anisotropic diffusion equation algorithm. The adaptive image denoising algorithm based on P-M diffusion equation is divided into two processes, the first is the process of adaptive edge detection and the other is the process of image denoising based on the result of edge detection. The specific behavior is that we improve diffusion coefficient of the original P-M diffusion equation and design an adaptive boundary detection function which divide the image into two areas at first. Then the diffusion eqution acts different models in different areas of image. By this way, we cannot only reduce the noise, but also protect the edge and ditails. At last, we use the structure tensor instead of gradient to detect the boundary of image so that we can improve the accuracy of boundary detection and reduce the impact of noise.In this thesis, different images are processed by the proposed method of adaptive Perona-Malik equation. The experiments prove that this algorithm is better than the traditional image denoising algorithms. It cannot only reduce the noise, but also protect the edge and details. By the subjective and objective analysis, it indicates that the adaptive partial differential equation algorithm is superior to the original partial differential equation algorithm.