Comic Image Segmentation and Adaptive Differential Evolution Algorithm with Different Times Characteristics: Based on Deep Plabv 3+

Abstract

Comic image segmentation involves dividing a comic image into distinct regions representing various elements like characters, backgrounds, and speech balloons to aid analysis, comprehension, and manipulation for tasks such as content retrieval, archiving, and production. However, accurately discerning between diverse visual elements, particularly in intricate or stylized artwork, poses a significant challenge. This manuscript proposes Comic Image Segmentation and Adaptive DE Algorithm with Different Times Characteristics (CIS-DTC-ADEA). Initially, the pictures collected from Comic Art information set are provided as input. The pictures are then sent into pre-processing. In pre-processing, Fast Resampled Iterative Filtering (FRIF) is used for noise removal, blur, and binary thresholding. After pre-process the output is fed to Deep Attention Dilated Residual DADRCNN for the classification of image. In general real and Synthetic Images was given for classification ADEA to optimize the Deep DADRCNN for classifying images us manga and classic. The proposed CIS-DTS-ADEA approach is applied in python working platform. A performance measures  of the suggested CIS-DTC-ADEA approach contains 20.28%, 28.22%, and 29.27% high accuracy, 16.23%, 28.21% and 22.35% higher kappa coefficient and 15.26%, 14.22%, and 22.07% high F-portion when analyzed to the current methods like Comic Unsupervised CD Technique for Heterogeneous RS pictures Based on Copula Mixtures and Cycle-Consistent Adversarial Networks (CUCD-HRSI-CCAN), Deep Learning-Based Classification of the Polar Emotions of “Moe”-Style Cartoon Pictures (CPE-SCP-CNN), High‑Precision Matching Algorithm for Multi‑Image Segmentation of Micro Animation Videos in Mobile Network Environment (MSMAVMN-HMA) methods respectively.