A New Method for Multilevel Thresholding of Crop Images Using Coronavirus Herd Immunity Optimizer

Due to the multimodality and uneven distribution of intensity levels in crop images, multilevel thresholding is a complicated job. In this paper, a new technique is proposed to segment the complex background color crop images (CBCCI) using recursive minimum cross entropy (R-MCE) and coronavirus herd immunity optimizer (CHIO). In the proposed method, CBCCI is converted into CIE lab color space, then pre-processed using Gaussian and Guided filters to smooth the flat part as well as preserve the color information on the edges. Finally, CHIO is applied with R-MCE to select the best possible threshold values. The accuracy of the proposed method is evaluated using peak signal-to-noise ratio, feature similarity index, structural similarity index, root mean square error, fitness value, and CPU time. To investigate the performance, a comparative study with bacterial foraging optimization, artificial bee colony, differential evolution, wind-driven optimization, firefly algorithm, sparse particle swarm optimization, and cuckoo search algorithm is made. The proposed method shows better average fidelity parameters than the above-reported algorithms. It also takes less computational time to obtain segmented images. Further, a graphical user interface is developed for consumer electronics applications which would be fast enough to process accurately and respond in real-time.