A differential evolutionary algorithm for multi-threshold image segmentation based on adaptive parameter control strategy

Abstract

Multi-threshold image segmentation is a simple and effective approach. Image segmentation techniques are significant in the fields of pattern recognition and computer vision. However, as the number of thresholds increases, the temporal complexity of selecting the best threshold increases exponentially. A meta-heuristic optimization approach called the differential evolution (DE) algorithm was utilized to address the optimization problem. This paper proposes an enhanced DE algorithm with adaptive control parameters (IJADE) for multi-threshold image segmentation. In this study, the enhanced DE algorithm optimizes five distinct thresholds for eight standard test images using the maximum between-class variance (OTSU) technique as the objective function. Comparison and analysis of IJADE with five other benchmark image segmentation algorithms demonstrated the viability and efficiency of the proposed method. The quantitative findings demonstrate that the peak signal-to-noise ratio and structural similarity index measure of the image segmentation results under various thresholds can be significantly improved by IJADE compared with existing benchmark methods. Peak signal-to-noise ratios for image segmentation for fabric and crane images were 22.197 and 23.1786, respectively, at a threshold of 5, with both placing top. With superior segmentation performance for digital image segmentation, IJADE was proven to be more effective.