Enhancing Image Retrieval Using Image Characterization with Adaptive Wavelet Transform

Abstract

With the exponential growth of digital multimedia, the demand for efficient image retrieval systems has become critical in domains such as healthcare, digital libraries, surveillance, and remote sensing. Traditional text-based retrieval systems, which rely on manual annotations, are limited by subjectivity, labor intensity, and scalability issues. Content-Based Image Retrieval (CBIR) addresses these limitations by extracting visual features such as color, texture, and shape directly from images. However, CBIR continues to face challenges such as the semantic gap between machine-computed features and human perception, variations in image quality, and computational complexity. To address these issues, this study proposes an adaptive wavelet transform framework guided by image characterization for enhanced retrieval accuracy. The approach leverages image characterization to analyze intrinsic properties such as texture orientation and energy distribution, enabling dynamic adaptation of wavelet decomposition. By tailoring feature extraction to individual image content, the method generates compact yet discriminative feature descriptors that improve retrieval performance across diverse datasets. Experimental evaluation demonstrates that the proposed model achieves robustness against noise, resolution variation, and compression artifacts, outperforming conventional static wavelet approaches. Applications in medical diagnosis, law enforcement, and multimedia management highlight the practical significance of this model. Ultimately, the research contributes to bridging the semantic gap and advancing the development of scalable, accurate, and context-aware image retrieval systems.