为了延拓灰度投影算法计算时间上的优越性能,解决传统投影法无法估计视频序列缩放参数问题,本文提出了一种新的基于尺度变化的灰度投影算法。算法根据缩放下投影矢量的特征,采用对数映射其坐标空间的方法,将矢量缩放参数降维成平移参数估计,再使用一维矢量互相关运算求解偏移量。实验结果表明,算法具有较好的鲁棒性,缩放系数位于0.7〈a〈1.3时,256×256图像块的解算时间为0.031s;参数求取精度〈0.03,能够较好的应用于视频帧序列的实时处理。
To extend the superior performance of gray projection algorithrn with low time cost to estimate the scale change of video sequence in real-time, a new gray projection algorithm based on scale change is proposed. According to the characteristics of the gray projection curves, the algorithm uses log transform to map the coordinate space of the curves, and the scale vector parameter is transformed into the translational parameter for estimation. Then, the offset is calculated with the 1-D vector correlation. The experiments show that, when the scale factor varying from 0.7 to 1.3, the computing time reduces from 1.220s to 0.031s with image block size 256×256, and the error is less than 0.03. This algorithm can be applied to the video real-time processing.