Studies On Segmentation Algorithm Of Animal Spermatozoa Image

With the development of the information technology, Image processing technology is becoming more and more important and powerful. In the field of medical image processing, it is an important application of image processing combining with the characteristics of medical images to achieve the process and analysis of image by computer.In the field of biomedicine image processing, image segmentation is the most important and difficult aspect. The segmenting quality has undeniable influence on quantitative analysis of cell information and cell aberrance. This paper deals with the research in the application of image process in the medical field. The main aim is to find an effective algorithm for animal's spermatozoa microscopic image segmentation by the image processing, which is prepared for quantitative analysis of spermatozoa information.After studying the technology direction and development of this field inside and outside our country, we summarize the medical image segmentation content and methods in detail, analyze the principle and characteristic of the present methods, present the advantages and disadvantages of these traditional image processing algorithms. Based on these theories and technologies, combining with the peculiarity of rabbit's micro-spermatozoa image-the uneven light of the background and low contrast, we present a new segmentation algorithm– the maximum between-class variance method based on edge detection. In order to make full use of the intensity space information, combining with the cell edges and 2-D grayscale distribution, an improved method 2D-Maximum entropy based on the characteristic of spermatozoa boundary. Experiments proves that our method possess good robustness and precision on special image.In the last chapter of the thesis, there is a detailed presentation about the basic theory of mathematical morphology. Then we introduce the watershed algorithm in detail. The author proposed an improved watershed image segmentation algorithm, which based on multi-scale grayscale morphology and gradient modification. Multi-scale morphological hybrid opening and closing by reconstruction is employed to smooth the original image, In order to eliminate over-segmentation problem, Multi-scale gradient modification is used before the standard watershed transform, Experiments showed that this method can efficiently avoid over-segmentation, and the segmentation results are greatly improved compared to the traditional algorithm.