点评：一种多视点图像的预处理方法，通过尺度不变特征变换算法在尺度空间对目标图像和源图像进行极值检测，提取目标图像和源图像关键点的尺度不变特征变换特征向量，利用特征匹配算法获得源图像和目标图像精确匹配的关键点对集合，并通过计算关键点对的乘性误差和加性误差对源图像进行颜色校正，通过在关键点对之间建立仿射变换关系，对颜色校正后的校正图像进行几何标定，优点在于本发明在保证多视点图像颜色校正和几何标定精确性的前提下，大大提高了颜色校正的鲁棒性和映射的精度，降低了几何标定的计算复杂度，提高了多视点图像的编码性能。

Comment: The invention discloses pretreatment method of view point image more than one kind, by the SIFT(scale invariant feature transform) method the metric space to the object map like looking like to carry out the extremum extracting with the source figure, extract the object map like with the source figure like the SIFT(scale invariant feature transform) features of key point vector, utilize the feature matching algorithm obtain the source figure like with the object map like the key point of exactly mating to the set, and look like to carry on the color calibration by calculating the right nature the taken advantage of sum of errors additivity error of key point to the source figure, by setting up the affine transformation relation in the key point between to, look like to go on geometry to the correction chart behind the color calibration and demarcate, the advantage lies in that the invention demarcates the on the premise of of accuracies with how much guaranteeing many view points image color calibration, has improved the robustness of color calibration and the precision of mapping greatly, has reduced the computive complexity of geometry demarcation, has improved the code performance of many view points image.