Wood Image Restoration And Edge Detection Based On Markov Random Field Theory

In recent years, the forest resources in our country face serious problem of shortage, the contradiction between supply and demand of wood is increasingly sharp. How to take use of and save the wood resources is a problem which is in badly need of solution. Applying the image processing techniques especially the technique of image restoration and edge detection to the research of wood images, is a main way to solve this problem. Through the restoration of damaged wood images and the edge detection of the wood outlines, we can supply technical support to the exploit and research of wood. This thesis apply the Markov random field theory to wood image restoration and edge detection, and realize better image restoration and edge detection.We first study the Markov random field theory, and analyzes the application of Markov random field theory to image restoration and edge detection. By combining the traditional techniques of image restoration and edge detection, we theoretically analyze the limitation of the traditional techniques, and highlight the applications and advantages of Markov random field theory in those two techniques.We utilize both the average filtering image restoration and the Markov random field theory image restoration to deal with the wood images with noises, and analyzes, under different choices of neighbor regions, the advantages and disadvantages of those two algorithms. Then by changing the type of the noises, we apply those two algorithms to the wood images with impulse noise and Gaussian noises separately, and analyze how those two algorithms depend on the type of noises. Then we judge, for different type of noises, whether the restoration effect of the Markov random field theory image restoration algorithm is always better than that of the average filtering image restoration algorithm.We utilize the edge detection algorithm to do the edge detection of original wood images. Because the annual rings of the wood are too many, the effect of edge detection is not so good. So we first apply Markov random field theory to strengthen the wood image, using high-frequency mapping and low-frequency mapping separately, and then do the edge detection of the strengthened wood images. We found that, for both the wood images strengthened by high-frequency mapping and low-frequency mapping, the effect of edge detection is pretty good. By changing the frequency regions and directivity of the edge detection, we repeat the above procedure and always get similar results.