Enhanced Video Shot Boundary Detection via KAZE Features Perceptual Hashing

Abstract

In the era of multimedia proliferation, efficient video shot boundary detec-tion is crucial. This study introduces a robust method leveraging KAZE feature descriptors and perceptual image hashing for accurate detection. Addressing challenges in multimedia data indexing and retrieval, our approach generates unique hashes for video frames, enabling precise transition identification. Experimental results on stan-dard datasets demonstrate superior performance, with a 94.2% average accuracy in detecting abrupt transitions, showcasing the method’s potential for multimedia sys-tems and digital communication applications.The proposed method is robust and discriminative, providing accurate results for short videos with less variation. The proposed method is robust and discriminative, providing accurate results for short videos. The efficiency and robustness of the technique are demonstrated through experiments on standard datasets such as TRECVid 2001 and TRECVid 2007. The RGB color histograms generated for detected cut transitions further validate the system’s performance. The full source code of the proposed approach (DOI link: https://github.com/hritxx/kaze-feature-descriptor-perceptual-image-hashing ) is uploaded to github, along with the associated data sets (readme files) to facilitate further research.