对全局运动和局部运动并存的序列低分辨率图像的配准，是超分辨率重建的难点之一。序列低分辨率图像间的非冗余信息会导致难以对不规则局部运动区域准确定位，降低配准精度。对此，设计了一种可自动区分运动类型并精确定位不规则局部运动区域的图像配准方法。首先，采用C均值聚类对运动属性分类得到全局运动配准结果；其次，分析非冗余信息的高频特性，采用高斯滤波器滤除其对不规则局部运动定位的干扰，提取局部运动区域；最后，设计自适应块尺度选择方法对不规则局部运动区域精确定位。实验结果表明：该方法可消除非冗余信息对配准的不利影响，提高配准精度。%Accurate registration of video sequences with both global and local motion is the challenging aspect of super-resolution reconstruction. The non-redundant information contained in multiple low-resolution frames will lead to inaccurate location and registration of the irregular local motion. This paper introduced a registration method enables to classify motion forms automatically and locate the irregular local motion accurately. Firstly, C means cluste- ring was used to classify the motion forms to obtain the registration results of the global motion. Secondly, the non- redundant information was removed by Gaussian filter. As the result, the local motion could be extracted. Finally, an adaptive block size adjusting method was designed to locate the irregular local motion accurately. Experimental results show that the proposed method can eliminate the interference of the non-redundant information during registration and improve the registration accuracy.
展开▼
