A Human Face Image Searching System Using Sketch

Thispaper addresses theproblem in mug shot searching in suspect database. The objective is to improve the eficiency of the suspect identificationprocess. The input is a sketch constructed by a witness. The output is the mug shot ranking for the suspect database. In the system, local and global feature measurements between sketch and mug shot faces are adopted. The local distance measurement compares facial primitives individually while the global distance measurement compares the geometric distances amongfeature points on theface. All mug shots are ranked based on the combined results. Experimental results show that 70% accuracy is achieved by considering the top 15% ranks.

• Type: Journal
• Year: 2002
• Where: MVA
• Authors: C. H. Man, Pong Chi Yuen
• Keywords: Computer Vision, Mug Shots, MVA 2002, Suspect Database, Thispaper Addresses Theproblem