Research On Image Segmentation Based On Wavelet Transform And Fuzzy Theory

Image segmentation is one of the basic and important technique in computer vision. Image segmentation is the process and related methods to segment an image into different sub-images with different characters and derive some interested objects. It is a key step from image process to image analysis, and it plays an important role in image engineering, and is also applied in a lot of fields such as computer vision, pattern recognition and medical image treatment. So image segmentation is a hot topic for researchers in the field of image process, which results in an amount of algorithms for image segmentation. However, it is discovered that currently available image segmentation algorithms applied to target detection and recognition, can't lead to satisfying results in the application and development. Thereby in this paper some research is carried out to improve the image segmentation algorithm.In this paper several aspects of image segmentation are discussed. Firstly, wavelet multi-scale edge detection algorithm is conducted based on some characters of wavelet transform such as low entropy, multi-resolution, decorrelation, selective multiplicity for basic wavelet and nicer time-frequency localization. When scaling vector is small, the precision of localization will be more accurate, but more noise will occur. Conversely, image edge in large scale will be more stable with less noise, but the precision of localization will be worse. Secondly, as the fringe part is fuzzy and others part is unambiguous correspondingly during the segmentation, multi-scale improved algorithm is proposed for wavelet edge detection. According to efficacious classification of fuzzy set membership function, ash image has been detected at multi-scale and its edge has been picked up by competition rule. Because the image boundary is fuzzy, edges are described further by idea of improved field. Then precise edges can be detected by algorithm. Aiming at some likelihood of color image segmentation which areas are scattered, unconnected, and over-segmentation, new color image segmentation algorithm is brought out based on feature divergence and fuzzy dissimilarity, which improves segmentation quality to some extent. Finally, we make good use of above algorithm, analyses results of images experiments from different fields, which prove that algorithm is efficient and feasible.