Abstract:

In the traditional maximum entropy threshold segmentation methods,the gray
probability of image is used and the corresponding gray value is ignored. In order to
adequately utilize the gray information and spatial information of the gray image,the
traditional 2D gray histogram is improved and the 2D Dvalue attribute gray histogram is
formed. Otherwise,the computation method for the average gray value and the 2D entropy is
improved.We use the spatial information value as a substitute for the gray probability to
compute the entropy.The computation of the entropy is based on the Dvalue attribute gray
histogram and creates the spatial different attribute information value entropy(SDAIVE). In
experiments,many different gray images are segmented with the improved maximum entropy
method and the traditional maximum entropy method,and the segmentation results are
compared.The experimental results show that improved threshold method can effectively
segment gray images and noise images. This method has strong antinoise capability and
clear segmentation results.

Key words: 2D histogram;entropy;gray information;image segmentation;gray probability