Automatic Ultrasound Image Segmentation by Active Contour Model Based on Texture

A novel automatic segmentation framework for ultrasound image processing, which is based on active contour model and texture similarity is reported. Textures are described by spectral measures based on the Fourier spectrum. Furthermore, correlation information is introduced to evaluate the similarity between textures which avoid handling vectored-value textures in the further procedure. An automatic reference feature selection scheme is also proposed. Based on these techniques, texture similarity map is introduced. Active contour model we use here incorporates region information, and the stop criterion is not based on gradients. We apply this model on texture similarity map and detect boundaries which are unclear in original intensity images. The experimental results show that our method is promising.