基于Harris角点全局运动估计的场景锁定算法

场景锁定技术是视频跟踪领域的一个关键技术，需要对图像的全局运动进行估计，常用的运动估计算法由于计算量大、对噪声敏感等因素很难得到实际应用。为了减少运动估计的计算量，提高全局运动估计的精度，提出了一种基于Harris角点全局运动估计的场景锁定方法。将图像分成4×4的16个块，选取每个块中响应值最大的角点，以参考图像角点周围矩形块与待匹配图像进行匹配，然后利用RANSAC算法对角点进行一致性检测，利用最小二乘法解算全局运动参数，最后计算图像之间的累积运动。实验结果表明，该算法运动估计精度高，稳定性好，能较好地实现场景锁定。

Scene Lock Algorithm Based on Global Motion Estimation of Harris Corner

Scene lock is an important technique in video tracking, which needs global motion estimation for images. Due tothe large amount of computation and noise sensitivity, traditional motion estimation algorithm is not suitable for effective ap- plication. In order to reduce the computation and enhance the accuracy of the global motion estimation, a scene lock algo-rithm based on global motion estimation of Harris corner is presented. The image is divided into 16 blocks of 4×4, and thecorner with the largest response in each block is selected. Then the surrounding rectangle of the selected corner is matched with the original image. Afterward, we use the RANSAC algorithm to verify the matching points, and use the least squaresestimation to solve the global motion parameters, and get the accumulative motion between images. The experimental results show that the posed algorithm is accurate and stable for motion estimation. Therefore, it can carry out the scene lock effec-tively.