An optimal color image multilevel thresholding technique using grey-level co-occurrence matrix

•This paper introduces gray-level co-occurrence matrix (GLCM) based color image Segmentation.•Cuckoo search (CS) is used in order to effectively enhance the optimal multilevel thresholding.•CS based GLCM was found to be more accurate for colored satellite image segmentation.•The feasibility of the proposed approach has been tested on various natural and satellite images. Image thresholding is a process that separates particular object within an image from their background. An optimal thresholding technique can be taken as a single objective optimization task, where computation and obtaining a solution can become inefficient, especially at higher threshold levels. In this paper, a new and efficient color image multilevel thresholding approach is presented to perform image segmentation by exploiting the correlation among gray levels. The proposed method incorporates gray-level co-occurrence matrix (GLCM) and cuckoo search (CS) in order to effectively enhance the optimal multilevel thresholding of colored natural and satellite images exhibiting complex background and non-uniformities in illumination and features. The experimental results are presented in terms of mean square error (MSE), peak signal to noise ratio (PSNR), feature similarity index (FSIM), structural similarity index (SSIM), computational time (CPU time in seconds), and optimal threshold values for each primary color component at different thresholding levels for each of the test images. In addition, experiments are also conducted on the Berkeley Segmentation Dataset (BSDS300), and four performance indices of image segmentation- Probability Rand Index (PRI), Variation of Information (VoI), Global Consistency Error (GCE), and Boundary Displacement Error (BDE) are tested. To evaluate the performance of proposed algorithm, other optimization algorithm such as artificial bee colony (ABC), bacterial foraging optimization (BFO), and firefly algorithm (FA) are compared using GLCM as an objective function. Moreover, to show the effectiveness of proposed method, the results are compared to existing context sensitive multilevel segmentation techniques based on Tsalli's entropy. Experimental results showed the superiority of proposed technique in terms of better segmentation results with increased number of thresholds.