Abstract
In this dissertation, the robust image matching methods by using gradient variations are studied. Two different approaches, with and without using the Hausdorff distance, for image matching methods are discussed in this dissertation. We examine our proposed methods in the face recognition, especially in different illumination conditions, since the gradient variations are very suitable for this area. Face image matching is an essential step for face recognition and face verification. It is difficult to achieve robust face matching under various image acquisition conditions. In this dissertation, a novel face image matching algorithm robust against illumination variations without using Hausdorff distance is shown. The proposed image matching algorithm is motivated by the characteristics of high image gradient along the face contour. We define a new consistency measure as the inner product between two normalized gradient vectors at the corresponding locations in two images. The normalized gradient is obtained by dividing the computed gradient vector by the corresponding locally maximal gradient magnitude. Then we compute the average consistency measures for all pairs of the corresponding face contour pixels to be the robust matching measure between two face images. To alleviate the problem due to shadow and intensity saturation, we introduce an intensity weighting function for each individual consistency measure to form a weighted average of the consistency measure. This robust consistency measure is further extended to integrate multiple face images of the same person captured under different illumination condition, thus making our robust face matching algorithm. Reliable image matching is important to many problems in computer vision, image processing and pattern recognition. Hausdorff distance and many of its variations have been employed for image matching with success. The second approach of the image matching method of this dissertation we proposed is an improved image matching method based on a modified Hausdorff distance with normalizing gradient consistency measure. This hybrid image matching combining Hausdorff distance with normalizing gradient matching is a new image matching algorithm integrates the geometric Hausdorff distance with the photometric intensity gradient information to obtain a better image similarity measure. To show the improvement of the proposed algorithm, we test it with some previous image matching methods on the problem of face recognition under lighting changes. Experimental results of applying the proposed face image matching algorithm with or without Hausdorff distance on the Yale face database or CMU PIE database are compared with those by previous matching methods. These results show superior recognition under different lighting conditions by using the proposed robust face image matching algorithm with or without Hausdorff distance.