利用统计概率的活动轮廓模型分割图像

Active contour model for image segmentation using statistical probability

活动轮廓模型分割图像依赖于图像目标区域和背景区域灰度的可分性,当图像结构复杂时分割结果往往不理想。可以利用图像变换得到图像特征向量,再研究图像特征向量和图像目标区域的隶属关系。把图像特征向量通过概率密度形成外力场集成到活动轮廓模型。这种改进的活动轮廓模型在图像分割中可以包含较多的图像分割信息,较强的抗噪声干扰性能,能对多种类型图像进行分割,同时它是图像分割活动轮廓模型的推广。对弱边缘区域、纹理图像、遥感图像的分割实验说明改进的活动轮廓模型有良好的分割性能和效果。

Active contour model for image segmentation depends on reparability of gray between the target region and the background. When the structures of images are complexity, the results of segmentation are often not ideal using active contour model. When image feature vectors are obtained by image transform, image feature vector affiliation with the target region are studied. The force field is formed by the probability density of the image feature vector. New force fields serve as external force field of active contour model This active contour model for image segmentation can contain more information of image, improve performance of anti-noise and extends types of image for image segmentation. At the same time it is the promotion of active contour models for image segmentation. Experimental results show that improved active contour model has good performance and effectiveness of image segmentation to weak edge region, the texture image and remote sensing image.