A novel adaptive contextual information-based 2D-histogram for image thresholding

This paper proposes a new two-dimensional (2D) histogram containing an image's adaptive local spatial contextual information. The correlation information between pixels is essential to achieve the optimal thresholds for image segmentation purposes. But it is an arduous task to find adaptive spatial information efficiently. So, an innovative adaptive local spatial information technique is proposed here to compute a novel 2D histogram. An adaptive contraharmonic mean (ACM) filter-based 2D-histogram is designed that contains contextual information by a unique proposed logical pseudo-energy (LPE) function. The order of contraharmonic filter Q is adaptively obtained using LPE, and the 2D Tsallis-Havrda-Charvat entropy is utilized to obtain the segmented results. Setting the Q value of the contraharmonic mean filter is tedious, and the proposed LPE function solves this problem. For the result assessments, the presented framework is compared with other maximum sum entropy-based methods, which are 2D-Kapur, 2D-Renyi, 2D-Tsallis, 2D-Masi and Parzen-window-based Kapur entropies-based image segmentation approaches. The proposed ACM-based Tsallis-Havrda-Charvat (Tsallis) entropy (ACM-Tsallis) shows excellent segmented results above local mean-based 2D-entropies methods and the latest Parzen-based Kapur entropy approach.