An image segmentation method based on two-dimensional entropy and variance
Juntao Xue, Zhengguang Liu, Xiuge Che · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2006
In this paper, we present a new image segmentation algorithm based on the concept of two-dimensional Renyi's entropy along with statistical variance from the assumed data sets of object and the background to produce the appropriate threshold. So the statistic infonnation, or relative spatial distribution, or co-occurrence, of pixel grey levels, was taken into account. Experimental results show that the method we proposed performed better than one-dimensional and two-dimensional entropy-based methods with lower segmentation errors, and a reduction in the amount of noise present in the resultant images. This method can be extended to any other entropy segmentation method based on two-dimensional gray histogram and may also be useful for pattern recognition and image sequence analysis. Especially when the gray value of the object and the background overlap greatly or there is big noises in the image, the segmentation result can be drastically improved.