Quantitative evaluation of image segmentation algorithms based on fuzzy convolutional neural network

Natural images are considered complex images in image processing because features such as color and texture are very much combined in these images. Generally, acceptable results are not obtained from the segmentation of this type of images therefore a better algorithm should be chosen from among the other algorithms. In this paper, a criterion based on the neuro-fuzzy approach is presented for the quantitative evaluation of image segmentation algorithms, which provides the segmentation evaluation of all types of complex natural images. In the proposed criterion, convolutional neural networks and fuzzy logic are used to bring the evaluation closer to human understanding. In this network, the similarity or dissimilarity of the obtained parts is trained to the network, and then in the test phase, the accuracy of the network output will be evaluated based on the proposed formula. The proposed method does not need a pre-segmented reference image and can be used in interactive and real-time systems. The obtained results show that the proposed method has shown a better efficiency compared to other criteria with an accuracy of 94.2%. © 2023, The Author(s), under exclusive licence to Bharati Vidyapeeth's Institute of Computer Applications and Management.