Region Based Segmentation by the Generation of Hierarchical Codebook Using Thermal Images

Abstract

Thermal images are playing a major role in a variety of applications like criminal tracking, land surveillance, human medical imaging to analyze the abnormal portions by the temperature variations thereby identifying the faults. Segmentation methods are used to retrieve the abnormal regions based on certain factors like intensity, colour, texture etc. To enhance processing time and segment thermal images more accurately, gray level RGB method is used. This paper discusses segmentation method based on region growing algorithm using super voxel by generating hierarchical codebook for thermal images. Thermal image is a visible image which is a distribution of different intensity levels where the image is segmented based on colour variations having codebook as the criteria. The obtained segments show better result to identify the abnormal regions. Using this algorithm segmentation of thermal images show accuracy up to 90% for the collected data set.