A new edge detection approach via neutrosophy based on maximum norm entropy

It is a quite important step to find object edges in applications such as object recognition, classification and segmentation. Therefore, the edge detection algorithm to be used directly influences the performance of these applications. In this study, a new edge detection method based on Neutrosophic Set (NS) structure via using maximum norm entropy (EDA-NMNE) is proposed. Many experts and intelligence systems, including the fuzzy system, do not satisfactorily succeed in resolving indeterminacies and deficiencies. However, in the NS approach, problems are solved by dividing them into True (T), False (F) and Indeterminacy (I) subsets. In addition, because the approach has a powerful algorithmic structure, NS's conditions with indeterminate and missing situations can be solved successfully. In this study, object edges can be found successfully with the proposed approach because edges of the object are considered as indeterminate. Thus, using the proposed EDA-NMNE, a strong intelligent expert system-based edge finding software was designed. In our study, 5 different object edge detection results were obtained by converting 5 different types of entropy into NS-based edge detection software. Thus, edge detection analysis has been performed by experimenting many types of entropy that have not been previously used for NS edge detection. The Metric Figure of Merit (FOM), Peak Signal-to-Noise Ratio (PSNR) analysis, mean square error (MSE) were used in this experimental study. Based on these results, highest FOM and PSNR values are obtained by using NORM entropy. In addition, the proposed edge detection approach were compared with Edge Detection Approach via Canny (EDA-C), Edge Detection Approach via Sobel (EDA-S), Edge Detection Approach via Variation-Adaptive Ant Colony Optimization (EDA-VAACO) and Edge Detection Approach via Active Contour without Edge (EDA-ACWE). FOM and PSNR tests have been used to evaluate the edge detection results obtained via 5 different methods. The findings demonstrated that the performance of the proposed edge detection approach is more successful compared to other methods.