Image Segmentation Base On Level Set And Its Application

The image segmentation is the most basic and important part of the digital image processing technology. It is the precondition of the most image analysis and image understanding methods. The image segmentation divided the image into several independent sub area by using different characteristic information, and make the homogeneous area shows the similar characteristics. Many theories and methods have been applied during the over the past few decades. And the image segmentation technology would be mature. In recent years,the image segmentation method based on level set is one of the most popular technology in the field of image segmentation. The method embed the initial curve into the level set energy functional which has the higher dimension as the zero level set. The zero level set function would evolve to the object contour by using the information of image statistics, and it can gain a natural contour by this way. At present, the image segmentation method based on level set is still in the stage of development, its theory and application aspects are subject to further research.The theory of image segmentation method base on the level set are studied in this paper.And two novel image segmentation method base on level set was proposed, then the method was applied in the field of food sampling and cell image segmentation. The specific works as follow:(1) This paper proposed the Adaptive Balance Energy Model base on Level Set(ABEM)as a sampling method about roast chicken. The global statistics information and the local statistical information of image are combined in the model. In order to improve the spreed and accuracy of image segmentation, the adaptive coefficient is used to adjust the balance of the them. Then, the energy of the image contrast is introduced to solve the problem of weak boundary image segmentation. This paper proposed the relative Index Model based on Color Satisfaction(CSIM) for detection of the chicken color quality. In the HSV color space, the core frequency domain of S component was computed as a color index of the roast chicken image, then the color satisfaction would be attained by using the value of color index to reflect the chicken color quality intuitively. This paper sampling the chicken regional from the combi-oven image and detection of color quality of the chicken by combined the ABEM and CSIM. They are available for monitoring the color quality of roast chicken. Also, they will be benefit for the measure-control system of the food quality in combi-oven.(2) The multiple adhere homogeneous area are often classified into the same area using the traditional level set method. The paper proposed a Multiphase Repulsive Level SetEvolution(MRLSE) model to solution the problem. The model make the Distance Regularized Level Set Evolution model as a energy term to guide the contour curve close to the target boundary. Then it uses the repulsive energy term which is proposed in this paper to maintain the independence of the multiple level set function. This proposed a novel method to segmentation the adhesion cells. The method have two procedure. First, the detection method base on the Circle Hough Transform(CHT) is used to confirm the center of cells. Then the MRLSE model will be used to extract the contour of the adhesion cells. Experiments show that this method has a high accuracy, and it can provide a important information for pathological analysis.