Forensic Identification Through Heterogeneous Sketch Recognition And Efficient Image Retrieval

The role of forensic identification for sketch recognition in criminal investigations becomes ever more critical since it is the only visual information concerning suspects provided by witnesses. To address the above issues simultaneously, this study proposes a new method by fusing heterogeneous sketch-face photo recognition and efficient image retrieval approaches for identification in forensic laboratories. This paper attempts to improve on matching sketches with photographic images using generative models such as the Generalized Gaussian Mixture Model (GMM) and content-based image retrieval (CBIR) methods, which include the Local Binary Patterns (LBP), Chain Code, etc. This system ensures efficient retrieval by employing novel techniques such as image fusion and statistical modeling. Experiments on the CUHK dataset show improved retrieval performance and up to 90% accuracy in Sketch-Based Image Retrieval (SBIR) systems. We present a novel method for large-scale face sketch–photo synthesis and find that it is effective in forensic applications, resulting in a significant acceleration of visual suspect identification from sketches.