Decoding comics: a systematic literature review on recognition, segmentation, and classification techniques with emphasis on computer vision and non-computer vision

The increasing popularity of digital comics in recent years has drawn substantial interest in comic recognition (CR). CR is a process that includes the automated recognition and extraction of comic panels, speech balloons, and text from comic images and covers a wide range of applications within the digital media sector. This study aims to thoroughly review the existing literature on CR methods and techniques to present all significant findings in recognition techniques, datasets, and comparative evaluations of recognition models. Systematic Literature Review (SLR) is conducted using a search strategy that includes various databases, and 60 studies from 2011–2024 are selected. Most of the studies examined for this study use computer vision (CV) techniques for CR, which includes 65% of studies. The remaining 35% of studies have used non-CV techniques. The studies employed segmentation techniques (40%), machine learning (28%), and deep learning (32%). Manga109 and eBDtheque are the most popular public datasets that 78% of the selected studies used. The remaining 22% of the studies built their datasets using existing datasets. The study summarises the findings of the CR research and emphasizes the need to establish uniform standards for accuracy and datasets. It highlights the necessity of investigating and creating hybrid approaches for effective CR operations.