中文摘要：

研究从多目标图像中自动提取单个目标的图像处理方法．从分析曲线的水平集表示人手，首先探讨了水平集动态轮廓分割和配准模型构建的统计思想和变分方法，然后针对多目标粘连图像的特点，提出了含边缘信息和先验形状的水平集图像分割模型，并将其应用于病原菌的识别．由于引入边缘信息改进对分割的约束，加强了目标边缘对分割轮廓的吸引，同时消除了一些由噪声、阴影和杂质造成的影响．实验表明，改进后的先验形状水平集图像分割方法能直接从多目标粘连图像中提取单个目标，进一步完善了依据显微镜图像识别病原菌的图像处理方法．

英文摘要：

Extracting one object from an image with multiple objects is researched. Level set segmentation and registration of contour lines were modeled with statistical and variational approaches, with level set based edges and shape prior segmentation model is constructed to deal with image with conglutinated objects, and applied in the segmentation of bacteria images. Edge information is added to regularization of segmenting line, such that attraction of object edges to active contour increases, and some effects led by noises, shadow and impurities of images are eliminated. Data show that improved level set based shape prior segmentation model can distinguish a single object from an image with several overlapped objects, and perfect the image processing for bacteria recognition based on microscope images.