Selection of Optimal Threshold in Digital Images using Collision Entropy and Min Entropy

Abstract

Abstract

Image segmentation partitions an image into multiple segments based on properties of discontinuity and similarity. Thresholding is one of the important methods for image segmentation. It is used to discriminate foreground from the background of an image. The selection of suitable threshold value in the image is a challenging task. Thresholding value depends upon the randomness of intensity distribution of the image. Entropy is a parameter to measure the randomness of intensity distribution of the image. In the proposed research work, Collision entropy based and Min entropy based approaches are proposed to select suitable threshold value. For Comparison, Shannon entropy based approach is also implemented. After this, threshold values obtained from five standard test images are evaluated using Peak Signal to Noise Ratio (PSNR) and Uniformity. From these experimental results, it is observed that Min entropy based approach is a better approach than Shannon and Collision entropy based approaches. 

Keywords: Image segmentation, thresholding, Shannon entropy, histogram, collision entropy, Min entropy

Cite this Article

Parmeet Kaur. Selection of Optimal Threshold in Digital Images using Collision Entropy and Min Entropy. Journal of Software Engineering Tools & Technology Trends. 2017; 4(2):     25–31p.