基于饱和度梯度的彩色序列图像光流场计算

传统的基于亮度守恒的光流法仅利用图像灰度信息进行光流求解,忽略了对包含丰富信息的色彩信息的利用。而彩色光流法虽然利用了彩色图像各通道信息,却忽略了对邻域信息的利用,造成存在各通道线性相关而无法求解的问题。同时,上述方法在光照变化时均不能得到正确的光流场估计。将基于饱和度的梯度计算方法引入灰度光流的计算方法中,提出了基于饱和度梯度的彩色光流计算方法,该方法有效地利用了彩色图像的色彩信息和邻域信息。试验结果证明：该方法提高了算法对光照变化的适应性,得到了更加精确的光流场估计,且避免了各通道线性相关的问题,为运动目标检测奠定了基础。

Optical Flow Field Computation of Color Image Sequence Based on Saturation Gradient

The traditional optical flow method based on brightness conservation only uses the image gray information for optical flow resolution,ignoring the application of the color information.While the color optical flow method utilizes the color image information from each channels,but neglecting the neighbor information,which leads to the linear correlation of channels but without any way of solution.Meantime,the methods mentioned above cannot obtain accurate optical flow field estimation under the condition of large varying illumination.The gradient computing method based on saturation is introduced to calculate the grey optical flow,and a new method is proposed to estimate optical flow field in a color image sequence,which makes good use of color information and neighbor information.Experimental results show that the presented method yields more accuracy optical flow field estimation under large varying illumination situation,avoids linear correlation of color components,and lays a good foundation for moving target detection.