Multi-level thresholding segmentation based on levy horse optimized machine learning approach

Image segmentation is considered one of the main image processing techniques based on image histogram examination. These methods are used to examine the image histogram and generate optimal thresholds to segment the images into regions by differentiating the thresholds. The thresholding technique is widely used in image segmentation due to its efficiency and accuracy. Image segmentation uses a multi-level thresholding technique called Otsu. The method loses accuracy when the thresholds are maximized due to high complexity and execution time. To overcome the issue, the proposed work uses a Levy horse-based support vector machine (LHSVM) to attain optimal multi-level thresholds with reduced error rate with an objective function Renyi entropy to return a fine-segmented informational image. Initially, the red, green, and blue (RGB) image is converted into grayscale and followed by the grayscale image is filtered using the non-local mean (NLM) filtering technique. Then, a 2D histogram is generated with the grayscale and filtered image. The 2D-histogram is given to the Levy Horse Optimization (LHO) to attain optimal threshold values. Finally, SVM is used to classify the image into foreground and background based on the optimal threshold condition and segment the image using the optimal threshold values. The image is segmented at different threshold levels, such as 2, 3, and 5. The performance metrics such as Peak Signal Noise Ratio (PSNR), Structural Similarity Index (SSIM), fitness, and CPU time are evaluated, and the proposed method outperforms other methods in terms of PSNR, SSIM, fitness and CPU time, respectively. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.