Facial image processing for facial analysis

This paper discusses a methodology for improved image processing for human facial analysis and tries to integrate results from the visible images with the corresponding thermal images. First, an enhanced face detection algorithm in color images is described. The performance of Haar Classifiers is known as a fast real-time face detection algorithm. However, it generates false detection. The suggested solution in previous research is to increase the training set. The proposed solution in this paper is to pre-process the color images by implementing color segmentation in Chrominance component and Hue component prior to face detection algorithm. Datasets from different resources are tested and the experimental results produced suggested that this approach increases the detection rate and reduces the false detection rate in some datasets, but not for all. The performances from these datasets are compared and the possible future implementation is discussed. Further, the detection is extended to eyes, nose, and mouth detection. The second contribution of this paper is to establish a link between the visible images and thermal image. The method of applying visible image information to locate the face features in thermal image is illustrated. Finally, some possible future implementation of the results of the facial analysis within the domain of security is discussed.