RESUMO
An uncontrolled lighting condition is one of the most critical challenges for practical face recognition applications. An enhanced facial texture illumination normalization method is put forward to resolve this challenge. An adaptive relighting algorithm is developed to improve the brightness uniformity of face images. Facial texture is extracted by using an illumination estimation difference algorithm. An anisotropic histogram-stretching algorithm is proposed to minimize the intraclass distance of facial skin and maximize the dynamic range of facial texture distribution. Compared with the existing methods, the proposed method can more effectively eliminate the redundant information of facial skin and illumination. Extensive experiments show that the proposed method has superior performance in normalizing illumination variation and enhancing facial texture features for illumination-insensitive face recognition.
RESUMO
A structural compensation enhancement method is proposed to resolve the issue of nonuniform illumination image enhancement. A logarithmic histogram equalization transformation (LHET) is developed for improving the contrast of image and adjusting the luminance distribution. A structural map of illumination compensation is produced with a local ambient light estimation filter. The enhanced image is obtained by nonlinearly fusing the LHET result, reflection component, and structural map of illumination compensation. Unlike existing techniques, the proposed method has the ability of two-way adjustment for brightness. Furthermore, the proposed method can effectively enhance the nonuniform illumination images with a balance between visibility and naturalness. Extensive experimental comparisons with some state-of-the-art methods have shown the superior performance of the proposed method.
RESUMO
Unique correct correspondence cannot be obtained only by use of gray correlation technique, which describes gray similar degree of feature points between the left and right images too unilaterally. The gray correlation technique is adopted to extract gray correlation peaks as a coarse matching set called multi-peak set. The disparity gradient limited constraint is utilized to optimize the multi-peak set. Unique match will be obtained by calculating the correlation of hybrid matrices consisting of reference differences and disparities from the multi-peak set. Two of the known corresponding points in the left and right images, respectively, are set as a pair of reference points to determine search direction and search scope at first. After the unique correspondence is obtained by calculating the correlation of the hybrid matrices from the multi-peak set, the obtained match is regarded as a new reference point till all feature points in the left (or right) image have been processed. Experimental results proved that the proposed algorithm was feasible and accurate.