Research On Morphology Watershed Methods For Medical Image Segmentation

Medical image segmentation is a key technology in the medical image processing andanalysis, and obviously good segmentation results can provide more important information todoctors and patients than ever for disease surveillance and correct diagnosis immediately. Aswell known, watershed transform segmentation technique is a widely used approach in themedical image segmentation technologies. However, the traditional watershed transformsegmentation algorithm has some advantages such as high computational efficiency and highaccuracy of segmentation, but it always lead to over-segmentation and noise sensitivity.Therefore, some morphological segmentation algorithms combing with the traditionalwatershed algorithm and some depth researches is proposed in this dissertation, the mainresearch works as follows:Cell image in getting often due to the interference of the external environment. Thisinterference will make cell image produce abnormal changes in texture, and making someunnecessary edges during the subsequent segmentation. And watershed transform algorithm isextremely sensitive to noise and gray change. In view of this phenomenon, we use a series ofmorphological operations to eliminate the false edge of the cell image and use gray differentialtransformation to suppress noise and uniform gray. This method effectively solve thewatershed algorithm in cell division over-segmentation and drain segmentation problem.Most cell adhesion and overlapping phenomena, which will make the segmentationalgorithm is difficult to accurately locate the edge of the overlap. Traditional watershedalgorithm basically reached the purpose of separation of overlapping cells. But it will increasethe number of separation points, and often over-segmentation cells. In view of this, we use themorphology algorithm to extract the geometric characteristics of the adhesion of cells, andthen using non-linear processing techniques to improve these geometric features, finally usethe watershed segmentation. This method will separate the adhesion cells more effectively.Most of the existing algorithms of cell image only divide adhesion or non-adherent cellimage. this algorithm combines the characteristics of a number of methods to improvemorphological watershed transformation. And draw a segmentation method can generally usedin adhesion or non-adherent cell image.The complexity of human tissue structure makes it uneven distribution of grayscale image. Because of the impact of these factors, the existing watershed algorithm is difficult toextract a meaningful part in the human tissues. In this paper, a large number of local minimumwhich have no values will be reduced by morphological algorithm. peak enhancement addedin images based on the fusion technology retains an important contours. This method guidancewatershed transform properly, and Get the right intuitive human tissue images.For each algorithm this paper chose some corresponding gray level image simulation inMatlab software. The experimental data derived results confirm the validity and accuracy ofthe algorithm.