Steganographic Embedding of Features through Hybrid Edge Detection for Effective Content based Image Retrieval
Prajakta Ugale, Poonam Ninad Railkar · 2025
Tremendous growth in social media platforms and online applications such as Facebook, Twitter, Instagram, YouTube, etc. is increasing multimedia data on the servers. As a result, content-based multimedia data retrieval based on machine learning has emerged as a demanding area to retrieve the meaningful content. The traditional Content based Multimedia Data Retrieval (CBMDR) stores the multimedia data and its extracted features separately which increases the storage requirement and adds complexity in indexing. There are numerous Content based Image Retrieval (CBIR) systems that considers ‘metadata’ or ‘keywords’ associated with images for image retrieval but it needs precise mapping of ‘metadata’ or ‘keywords’ with the images. This paper presents a memory efficient CBIR that uses stenographical approach to embed features in the image itself using hybrid edge detection to reduce the memory footprints and indexing overhead during retrieval. The effectiveness of the suggested approach is assessed based on Oxford 102 flower and Caltech 101 Object dataset providing retrieval accuracy of 95.10% and 97.62% respectively, and it helped to diminish the complexity in retrieval as compared with previous approaches.