Image Segmentation Recommender Using Bio-inspired Algorithms
Veni Devi Gopal, G. Shreedevi, Angelina Geetha · International Journal of Electronics and Communication Engineering · 2024
An important aspect of image processing is image segmentation. It has been applied to a wide range of tasks, including augmented reality, video surveillance, item detection, and medical picture analysis. Even though many different algorithms have been created for picture segmentation, it has never been easy to determine which approach is appropriate for a particular image. Since every image has unique characteristics, it is impossible to find an algorithm that works for every image. Thus, one of the most difficult tasks is determining which method is best for a certain image. In earlier research, we adapted three meta-heuristic clustering algorithms (Shuffled Frog Leap, firefly, and spider monkey algorithm) and demonstrated their superior performance over the widely used k-means technique. It has also been demonstrated that the three algorithms were able to get over the main drawback of the k-means algorithm, which is its automatic determination of the “k” value. The goal of this work is to create a recommendation system that can identify the optimal segmentation algorithm based on the image input within a minimal time span. The outcomes demonstrated that, in terms of SSIM, FSIM, and time required, the suggested method is capable of recommending the best segmentation algorithm. It was found that the suggested algorithms required less computing time and were more than 90% efficient. The Open Surfaces dataset, the Berkeley Segmentation Dataset and Benchmark (BSDS) were used to test the recommender system.