Applied learning-based color tone mapping for face recognition in video surveillance system

摘要

In this paper, we present an applied learning-based color tone mapping technique for video surveillance system. Thistechnique can be applied onto both color and grayscale surveillance images. The basic idea is to learn the color or intensitystatistics from a training dataset of photorealistic images of the candidates appeared in the surveillance images, and remapthe color or intensity of the input image so that the color or intensity statistics match those in the training dataset. It iswell known that the difference in commercial surveillance cameras models, and signal processing chipsets used by differentmanufacturers will cause the color and intensity of the images to differ from one another, thus creating additional challengesfor face recognition in video surveillance system. Using Multi-Class Support Vector Machines as the classifier on a publiclyavailable video surveillance camera database, namely SCface database, this approach is validated and compared to theresults of using holistic approach on grayscale images. The results show that this technique is suitable to improve the coloror intensity quality of video surveillance system for face recognition.© (2012) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.