Choosing parameters for Rényi and Tsallis entropies within a two-dimensional multilevel image segmentation framework

In this work, the effect of Rényi and Tsallis entropies' parameters on the image segmentation quality within a two-dimensional multilevel thresholding framework is assessed and analyzed. The problems of automatically tuning entropy's parameter and determining the optimal thresholding values are solved in a single task. This is done by using the Quantum Genetic Algorithm (QGA). The numerical experiments conducted on different types of images demonstrated that Rényi and Tsallis entropies perform approximately similarly, and they are optimal when their parameters are null. Moreover, it was shown that optimizing the entropy does not lead to maximize the Peak Signal to Noise Ratio (PSNR) and the Structural SIMilarity (SSIM) criteria. Then, we have proved that these two criteria are not sufficiently consistent with human visual perception. Finally, the comparative study performed on some synthetic and real images demonstrated the effectiveness of the proposed method.