Showing papers by "Malay K. Kundu published in 2020"

Interval Type-2 Fuzzy Set and Theory of Weak Continuity Constraints for Accurate Multiclass Image Segmentation

Abstract: Multiclass image segmentation is a challenging task due to the uncertainties involved with the process of segmentation. To handle those uncertainties, we propose an automatic multiclass image segmentation method based on an interval type-2 fuzzy set (IT2FS). In the proposed method in this article, the accurate multiclass segmentation is achieved by minimizing an energy function. This energy function is based on IT2FS and weak continuity constraints present in the membership values. The theory of weak continuity constraints helps to localize the segmentation boundaries between the classes accurately with the minimization of the energy. The proper localization of segmentation boundaries helps to minimize the uncertainties in the segmentation process. We also theoretically show that the minimization of the energy function reduces the uncertainties present in the segmentation process. Furthermore, the method automatically determines the number of clusters without a priori knowledge. The proposed method is found to be superior to the existing conventional, fuzzy type-1 and fuzzy type-2 based segmentation techniques. The superiority is verified using synthetic and benchmark datasets. The noise immunity of the proposed method is found to be better than that of the state-of-the-art methods when benchmark against the modified Cramer–Rao bound.

Multi-class Image Segmentation Using Theory of Weak String Energy and Fuzzy Set

Abstract: Segmentation of a multi-class image is a major challenging work in image processing. The challenge arises as the uncertainties occur in the segmentation process. Here we present a novel method based on the concept of weak string energy to manage the uncertainties in the segmentation process. The concept of the weak string is utilized to find the location of the boundaries accurately among the segments. The segments of an image are generated based on the energy function in the fuzzy set domain in the proposed method. The accurate segments are generated when the function attains its minimum value. The segments are generated from an image without any prior knowledge about the total count of segments. The performance of the method is verified experimentally using different datasets and it is found to be quite satisfactory compared to the state-of-the-art methods.

Nondestructive Testing Image Segmentation based on Neutrosophic Set and Bat Algorithm

Abstract: Industry uses nondestructive testing (NDT) to detect a fault in metal without damaging it. Image segmentation based technique for detecting the fault from an NDT image is a difficult task. The difficulty emerges due to uncertainties in the NDT image pattern. To segment an NDT image efficiently the uncertainties should be handled efficiently. In this paper, we present a novel technique to segment an NDT image by handling the uncertainties based on neutrosophic set(NS). The NS manages the uncertainties by representing an image into a true, false, and indeterminate subset. For proper NS value representation, two operations α – mean and β – enhancement are essential. For finding the proper values of α and β depending on the image statistics we utilize the bat algorithm(BA). The algorithm finds the optimal values of α and β for managing the uncertainties properly. We find that in terms of performance the proposed method is quite satisfying in comparison to the latest methods.