Dominant color component and adaptive whale optimization algorithm for multilevel thresholding of color images

Optimal multilevel thresholding for image segmentation got much importance in recent years. Several entropic and non-entropic objective functions with evolutionary computing algorithms have been successfully implemented to get the optimal multilevel thresholds for gray scale images. The problem of multilevel thresholding becomes complex for color images. Because, the basic color components (red, blue, green) of the color image are extracted and the multiple optimum threshold values are calculated for each of the components separately. This makes the methods computationally intensive and inaccurate. Further, the required color information is not retained in the thresholded output. To solve these problems, an efficient technique is proposed in this paper, extracting only the dominant color component (DCC) of an image, for optimal thresholding. A novel segmentation score is introduced to justify the methodology. The optimum threshold values are obtained using a newly suggested evolutionary computing technique named adaptive whale optimization algorithm (AWOA). The main contributions are - (i) a novel DCC approach is introduced, (ii) an efficient optimizer AWOA is proposed, (iii) a new segmentation score is introduced, (iv) experimental results on standard test color images are explored. The outcomes are compared with all existing method's approaches (using all the RGB components) on color image thresholding. Its performance analysis using standard metrics is deliberated in detail. Statistical analysis is also performed. From the outcomes, it is perceived that the suggested DCC-AWOA concept yields high quality segmented images. The work may encourage further research to explore its high dimensional applications. (C)& nbsp;& nbsp;2022 Elsevier B.V. All rights reserved.& nbsp;

Automated summary

Optimal multilevel thresholding for image segmentation has gained significant importance in recent years. Various entropic and non-entropic objective functions combined with evolutionary computing algorithms have been successfully applied to obtain optimal multilevel thresholds for grayscale images. However, multilevel thresholding becomes complex for color images and the methods used are computationally intensive and inaccurate. In this paper, a novel technique is proposed that extracts only the dominant color component (DCC) of an image for optimal thresholding. A new segmentation score is introduced to evaluate the methodology. The optimal threshold values are obtained using the adaptive whale optimization algorithm (AWOA). The main contributions of this work are the introduction of the DCC approach, proposing an efficient optimizer (AWOA), introducing a new segmentation score, and conducting experiments on standard color images. The results demonstrate that the suggested DCC-AWOA concept produces high-quality segmented images. This work may inspire further research in high-dimensional applications.