Abstract: The foreground image is hard to be extracted when image’s intensity is inhomogeneous or the contrast is low.To solve this problem,an image segmentation method was proposed.This method reconstructs background model by combining sine basis functions with absolute distance measurement,and solves the model according to optimization theory and iteration method.This method discriminates between background and foreground of every pixel by comparing the intensity of each pixel in the background model with that in real image.To deal with the situation of inhomogeneous intensity of image,the image is divided into blocks before image segmentation,and is segmented according to background model and background similarity among adjacent blocks in sub-block image.Experimental results show that,comparing with classical methods including fuzzy C-means and OTSU,this method has lowest segmentation error,especially for the image with inhomogeneous intensity and low contrast.In the applications of palmprint image segmentation,comparing with iterated line tracing method and rough-fuzzy set method,this method has the characteristics of low error rate,high signal to noise ratio and shorter processing time.

Key words: Background model,Image segmentation,Basis function,Sine function,Optimization theory,Iteration method,Distance measure,Face recognition