提出一种由粗到精的透视畸变最小化算法,借助大位移视图来修补目标图像.它先对大位移视点图像进行透视畸变校正后再用来补全目标图像上的丢失信息区域.首先,在平面场景的假设下,大位移视点图像通过单应矩阵进行全局变形得到初始的畸变校正.然后,由误匹配识别机制检测出初始校正的大位移视图中的残余畸变.在颜色一致性和位移场光滑性的期望下,残余畸变通过基于能量优化的重叠像素对应算法得到进一步的松弛.最后,在极线几何以及像素邻域中的位移场光滑性和颜色一致性的约束下,信息丢失像素按照特别定义的修补优先级函数依次得到恢复.泊松图像融合算法被用于消除修补区域与其周围像素之间可能存在的鬼影现象从而得到无缝的修补效果.实验表明,该方法优于已有的图像修补算法,且能够修补含有复杂结构信息的较大受损区域.

A coarse-to-fine perspective distortion minimization algorithm is proposed for image repairing based on an additional large displacement view (LDV) of the same scene. It works by correcting the perspective distortion in the LDV image, and then utilizing the rectified LDV image to recover the missing areas on the target image. First, under the assumption of a planar scene, the LDV image is globally warped according to a homography to generate the initial distortion correction. Second, a mismatch recognition mechanism detects the remaining distortions in the initially corrected LDV image. They are further relaxed by energy optimization of overlap correspondences with the expectations of color constancy and displacement field smoothness. Third, under the constraints of epipolar geometry, displacement field smoothness and color consistency among the neighboring pixels, the missing pixels are orderly restored according to a specially-defined repairing priority function. Poisson image blending is adopted to eliminate the ghost effect between the repaired region and its surroundings and get the seamless repairing effect. Experimental results demonstrate that this method outperforms recent state-of-the-art image completion algorithms, especially for completing large damaged area with complex structure information.

Supported by the National Natural Science Foundation of China under Grant Nos.60403038, 60603076, 60703048 (国家自然科学基金); the National High-Tech Research and Development Plan of China under Grant No.2007AA01Z316 (国家高技术研究发展计划(863)); the National Basic Research Program of China under Grant No.2002CB312101 (国家重点基础研究发展计划(973)