Biologically-inspired pre-compression enhancement of video for forensic applications

Abstract

Digital video is an increasingly common form of legal evidence. Visible artifacts caused by lossy compression raise questions of the validity and therefore admissibility of such evidence. Blockiness (caused by the loss of high frequency spatial information) and translation of artifacts (due to motion-based compression) are two major sources of distortion which can affect object shapes and therefore object recognition, such as that of alphanumeric characters. We propose a novel pre-compression video enhancement method inspired by the properties of Large Monopolar Cells (LMC) in insect visual systems. The integration of standard compression schemes and biologically-inspired video enhancement enables us to keep more semantic information at comparable data rates, thereby improving recognition rates. Experimental results demonstrate the effectiveness and limitations of the proposed approach.