Research Of Image Segmentation Based On Fuzzy Mathematical Morphology And Watershed Algorithm

Image segmentation is the key step from image processing to image analysis, and also is a basic computer vision technology. This is because the image segmentation, target separation, feature extraction and measurement parameters make the original image into a more abstract and more compact form, making more high-level analysis and understanding possible. Therefore, over the years people have paid much attention on image segmentation. Although there are many classic segmentation algorithms, we still don't have a segmentation method which can make all the images have ideal segmentation results. According to the different characteristics of image segmentation and known priori knowledge studying the segmentation model of specific image characteristics is an important way to raise image segmentation.In image processing, image structural characteristics are very obvious. If we seize this feature, we will reduce a large number of the processing time in image processing. Because morphological techniques in image processing take the structural characteristics of the image fully into account, compared to other image processing methods, the technology has the unique morphological characteristics of the structural advantages. Morphological watershed transform is a mathematical morphology image segmentation method. It can access an accurate verge. That is continuous, closed single-pixel width verge. It has been widely used in many fields of image processing. But its main shortcoming is that it is very sensitive to noise, the weak noise will cause serious over-segmentation phenomenon. It resulted in bringing a large number of separate regional fragmentations which is an over-segmentation phenomenon. In this paper we process images in two phases to solve the segmentation problem, can reduce regional segmentation effectively. The research work includes the following aspects:1) In Chapter 2 and 3, We introduce the fuzzy morphology, watershed algorithm theory and concrete steps of the algorithms, describing from the definition of algorithm to the algorithms' concrete steps. 2) Based oh the depth research and analysis of image forest transform (IFT) watershed algorithm, against the over-segmentation phenomenon we propose an improved watershed algorithm. Firstly, we use automatic identification of the optimal threshold method to determine the image segmentation threshold. Secondly, we modifies the constraint of IFT watershed algorithm path cost function appropriately, and proposes an improved algorithm to reduce the IFT watershed algorithm storage space and speed up the implementation speed. The experimental results show that the algorithm achieved the complete separation of the objectives and background, and retained complete details at the same time.3) Based on the above-mentioned work and combining the advantages of fuzzy algorithm and the watershed, this paper proposes a combined image segmentation method of the two methods. We use fuzzy image morphology to smooth the image, and then use improved IFT watershed segmentation algorithm to segment the image. Finally we have the final segmentation results. The method resolves over-segmentation and makes the follow-up treatment complete more quickly.4) To testify the effectiveness, we apply this method proposed in the paper to medical images and agriculture images through Matlab. The experiment results show that the method could resolve the over-segmentation and edge discontinuity, and achieve better segmentation results. Finally, Chapter 5 summarizes the main work of the full thesis and points out the prospect of the farther direction.