Research On Image Filtering And Segmentation Of Edge Preserving In High Intensity Noise

In recent years, digital image processing technology has been widely used in industrial, medical, aerospace and other fields, and the specific application includes robot view, cell classification, face recognition, remote sensing image analysis, and so on. Due to various reasons, images are often contaminated by noise during the process of acquisition and transmission, and thus leading to decreasing the performance of the subsequent image processing algorithms. Therefore, it is urgent to solve the problem of image processing under the condition of high intensity noise.This paper launches the research on image filtering and image segmentation for high intensity noise. It tries to design an effective image filtering algorithm and image segmentation algorithm, in order to remove the noise and while keeping the edge details of the image as much as possible. Specific research contents include:(1) The existing image filtering algorithms are often unable to keep the details of the image and suppress the noise at the same time. In this paper, an image filtering algorithm based on alpha truncation and edge preserving is proposed. The algorithm at first uses alpha-trimmed method to form a guide image, and then establishes the local linear model between the image and the target image.In this way, the weights which have the property of edge preserving are constructed. This algorithm can not only effectively remove many kinds of noise, but also can preserve the edge information. Experiments on both artificial and real images show the effectiveness of the proposed algorithm.(2) The existing fast image segmentation algorithm based on the fuzzy C means cannot directly used to handle color image. In this paper, the fast image segmentation algorithm is proposed for the color image with high intensity noise. In this algorithm, the similarity between neighbor pixels and the center points is defined according to the color space, and by using this similarity the image color for each pixel has been corrected. On the basis of this, RGB color space is reduced from 256* 256* 256 to 216. Finally, fast FCM is used to segment the color image after quantization. The experimental results are tested on artificial images and real images. The experimental results show that the algorithm has higher accuracy and lower time complexity.(3) Non-local image segmentation method is used to identify pixels with the same structure properties by measuring the similarity between image blocks, so the time complexity of the method is high. Local image segmentation uses local spatial neighborhood pixel information to affect the classification of each pixel, so this kind of method ignores the effects of the non-local pixels. Aiming at these problems, an image segmentation method based on both non-local and local similarity is proposed. This algorithm combines two parts together, and designs a new weight. And the proportion of the two parts is controlled by an adaptive parameter. Then, the fast FCM is used to segment the gray image. The experiments are performed on artificial images and real images, and the corresponding results show that our algorithm is better.