Image thresholding using Tsallis entropy

Image analysis usually refers to processing of images with the goal of finding objects presented in the image. Image segmentation is one of the most critical tasks in automatic image analysis. The nonextensive entropy is a recent development in statistical mechanics and it is a new formalism in which a real quantity q was introduced as parameter for physical systems that present long range interactions, long time memories and fractal-type structures. In image processing, one of the most efficient techniques for image segmentation is entropy-based thresholding. This approach uses the Shannon entropy originated from the information theory considering the gray level image histogram as a probability distribution. In this paper, Tsallis entropy is applied as a general entropy formalism for information theory. For the first time image thresholding by nonextensive entropy is proposed regarding the presence of nonadditive information content in some image classes. Some typical results are presented to illustrate the influence of the parameter q in the thresholding. (C) 2004 Elsevier B.V. All rights reserved.