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

Segmentation for Multimodal Brain Tumor Images Using Dual-Tree Complex Wavelet Transform and Deep Reinforcement Learning

“…As can be seen from Figure 5 , after the number of pixel points is increased to 200, the image similarity value of the proposed algorithm is always higher than that of other literature algorithms, and it shows significantly higher than that of other literature algorithms at each pixel point; especially, when the number of pixel points is 500 and 3000, the similarity curve of the proposed algorithm shows two small peaks with significant advantages. In contrast, the similarity values obtained with the other algorithms are lower than those of the original image, in which, the similarity of Yang et al [ 6 ] is relatively high above 0.7, the similarity of Dhar and Kundu [ 7 ] is also close to 0.7, the highest similarity of the algorithms of the Dissanayake et al [ 8 ] and Zhou et al [ 9 ] is above 0.6, and Sun et al [ 10 ] has the lowest and does not exceed 0.6. The feature quantity of these algorithms differs greatly from the original image, indicating that the images segmented by the algorithms do not retain the original information well and are less effective.…”

“…With the increase of image information, the highest information loss index of the proposed algorithm is only 0.18. However, the information loss index of the algorithms of Yang et al [ 6 ], Dhar and Kundu [ 7 ], Dissanayake et al [ 8 ], and Zhou et al [ 9 ] reaches 0.30 when the amount of image information is 500; the information loss of the algorithm of Sun et al [ 10 ] is severer, and the information loss index is 0.37 at an image information level of 500. As shown in Figure 6 , the information loss index curves of the proposed algorithm are lower than other literature algorithms at each image information amount, which has a significant advantage.…”

“…CT images of brain tumors from 50 cases were selected in BraTS2018 dataset, CQ500 dataset, and a hospital brain tumor dataset, respectively, to extract image information; 70% of the data are used for algorithm training and 30% of the data are used for algorithm testing. CT images of brain tumors in three datasets were used as experimental objects for detailed segmentation of multimodal tumor images, and were compared and analyzed with Yang et al [ 6 ], Dhar and Kundu [ 7 ], Dissanayake et al [ 8 ], Zhou et al [ 9 ], and Sun et al [ 10 ] to determine the effectiveness of the proposed algorithm.…”

“…However, this algorithm does not consider noise points in the original image, which leads to low accuracy of subsequent segmentation. Dhar and Kundu [ 7 ] used a segmentation algorithm based on fuzzy set with weak continuous constraint theory; the fuzzy feature set was initially established based on the features of pixel points in the image, and the segmentation was performed based on the pixels with consistent correlation by combining the theory that the pixels with consistent feature expression in the image generally have weak correlation phenomenon. However, this algorithm ignores the diversity of pixel points in the initial feature determination of the algorithm, which leads to low accuracy and affects the subsequent segmentation accuracy.…”

Interval type-2 fuzzy logic system based similarity evaluation for image steganography

“…Hence, a significant improvement has been made in shifting from a type-1 fuzzy logic system (T1 FLS) to interval type-2 fuzzy logic system (IT2 FLS) by the researchers in recent years. IT2 FLSs have been effectively applied in the various applications of image processing systems such as classification (Majeed et al, 2018;Rubio et al, 2017), filtering (Singh et al, 2018), segmentation (Dhar and Kundu, 2019;Zhao et al, 2019), and edge detection (Castillo et al, 2017;Gonzalez and Melin, 2017;Gonzalez et al, 2016;Martínez et al, 2019;Melin et al, 2014).…”