Research And Application Of The Edge Detection In Noise Image With Fuzzy Objects

As an important area of image analysis, edge detection plays an important part in the area of pretreatment; hence the edge detection technology has recently become popular in recognition. However, in practice, as good or bad quality of the image and complexity of the target, therefore edges are usually extracted from enhanced and noise-reduced image. In addition, with the development of the signal processing, fuzzy mathematics, geometry and other infrastructure, more and more techniques are used in the image processing. So many algorithms are presented such as wavelet, grey forecasting model, fuzzy, genetic algorithm neural network, morphology, fractal theory and so on. It is difficult to detect the edge of these fuzzy targets in noisy image. The paper mainly solves this problem by the way such as the noise-reduced, fuzzy edge enhancement, target detection and segmentation, object edge extraction, fractal theory, cluster analysis. The main contributions of this thesis include following issues:1. The image processing technology of the edge detection was introduced. The traditional edge detection algorithm and edge detection based on the application of new technologies were presented, and summarized the basic viewpoints of edge detection and basic research approach;2. The edge detection algorithm made the noise become big, so the improved edge detection algorithm which based on fuzzy entropy and fractal dimension was proposed. It used fuzzy entropy to suppress the noise increased, and used the fractal dimension to describe the image local characteristics. The experimental results show that the algorithm can suppress the noise expanded to obtain better edge features in salt-pepper noise image;3. The multilevel fuzzy enhancement edge detection algorithms enhanced some edges at expense of the weakening of other edge, so a multi-level fuzzy enhanced edge detection algorithm was presented. First, the Valley-emphasis algorithm is employed to estimate the optimal threshold parameters. Then, the new convex non-linear membership function based on this threshold is defined to map the fractal gray image into fuzzy feature plane. Finally, fuzzy enhancement with separated regions and smooth processing are given. On this basis, the fuzzy entropy measure is employed to extract edge. Comparing experimental results show the feasibility and effectiveness of the algorithm; 4. In noisy image, as noise pollution made the histogram valley no longer obvious and the initial value of random target selection results instability in the clustering algorithm. Through a combination of pixels in the spatial information, the detection algorithm based on two-dimensional histogram of cluster was presented. The algorithm introduced the Otsu algorithm to determine the cluster center of the initial value of the two clustering and fuzzy cellular automata to determine the edge of the clustering segmentation. Experimental results show that the algorithm can extract the correct and continuous edge, and detection results of the target are stable.In a word, we have made a lot of fruitful attempts and significant progress on recognition of the multi-objective of which one or more is week contrasting to others.