Joong Kyu Kim

A new method of illumination normalization for robust face recognition

In this paper, we propose a novel method of illumination normalization developed on the basis of the retinex theory. In retinex based methods, illumination is generally estimated and normalized by first smoothing the input image and then dividing the estimate into the original input image. The proposed method estimates illumination by iteratively convolving the input image with a 3×3 averaging mask weighted by an efficient measure of the illumination discontinuity at each pixel. In this way, we could achieve a fast illumination normalization in which even face images with strong shadows are normalized efficiently. The proposed method has been evaluated based on the Yale face database B and the CMU PIE database by using PCA. Carrying out various scenarios of test, we have found that our method presented consistent and promising results even when we used images with the worst case of illumination as training sets. We believe that the proposed method has a great potential to be applied to real time face recognition systems, especially under harsh illumination conditions.

Common image method(null space + 2DPCAs) for face recognition

In this paper, we present a new scheme called Common Image method for face recognition. Our method has a couple of advantages over the conventional face recognition algorithms; one is that it can deal with the Small Sample Size(SSS) problem in LDA, and the other one is that it can achieve a better performance than traditional PCA by seeking the optimal projection vectors from image covariance matrix in a recognition task. As opposed to traditional PCA-based methods and LDA-based methods which employ Euclidean distance, Common Image methods adopted Assemble Matrix Distance(AMD) and IMage Euclidean Distance(IMED), by which the overall recognition rate could be improved. To test the recognition performance, a series of experiments were performed on CMU PIE, YaleB, and FERET face databases. The test results with these databases show that our Common Image method performs better than Discriminative Common Vector and 2DPCA-based methods.