特征描述器在许多机器视觉任务中有着重要的应用。为了提高特征描述器的性能，在极坐标下设计出了一种新的特征描述器——小波差分统计描述器(harr wavelet differential histograms，HWDH)。该描述器算法先把选定的特征区域分成9个小区域；接着分别对小区域进行小波变换；然后算出对应的方向差分，同时用改进的尺度适应性二阶矩矩阵来选择特征点的主方向；最后在极坐标下，对每个小区域的差分，在相应的方向条目内进行统计。与其他基于统计的描述器相比，该算法不需要对特征区域进行旋转，即可减少统计误差。经与尺度不变特征变换（SIFT）的描述器的实验比较表明，该描述器在鲁棒性和速度性能上都很好。

Descriptor of features point is vital application of many computer vision tasks. Therefore, this paper presents a novel algorithm to design the descriptor of image feature points based on differential histograms under polar coordinates (HWDH).Firstly, the neighbor area of feature point is divided into nine small patches. Then, Gradient differential is produced by harr wavelet on every small patch and a dominant direction of the feature point is computed by Scale Adapted Harris detector which is improved. Lastly, histogram is generated from the differential statistic. In contrast to other descriptor based on differential histograms， the description generated by our algorithm needn’t to rotate in the around area of feature point. Therefore, it can deduce the error from statistic. Moreover, the comparative experiments illustrated that the proposed algorithm is more rapid and accurate than SIFT.