A comparative study on Kapur's and Tsallis entropy for multilevel thresholding of MR images via particle swarm optimisation technique

Abstract

The present paper explores both the Kapur's and Tsallis entropy for a three level thresholding of brain MR images. The optimal thresholds are obtained by the maximisation of these entropies using a population-based search technique called as particle swarm optimisation (PSO). The algorithm is implemented for the segregation of various tissue constituents, i.e., cerebral spinal fluid (CSF), white matter (WM) and grey matter (GM) region from the simulated images obtained from the brain web database. The efficacy of the thresholding methods was evaluated by the measure of the spatial overlap, i.e., the Dice coefficient (Dice). The experimental results show that: 1) for both the WM and CSF the Tsallis entropy outperforms the Kapur's entropy by achieving an average value of 0.967279 and 0.878031 respectively; 2) for the GM, the Kapur's entropy is more beneficial which is duly justified by the mean value of Dice which was 0.851025 for this case.