一种基于马尔科夫链模型的运动估计新方法

A new method of motion estimation algorithm based on markov chain

运动估计是视频编解码系统中的关键技术,而运动估计的快速块匹配算法主要是通过采用初始搜索点预测,提前退出技术以及不同的搜索模板来提高算法的效率。通过引入马尔科夫链模型实现了对初始搜索点的准确预测,以及利用模糊逻辑和遗传算法以避免搜索过程陷入局部最优点等方法,提出了一种基于马尔科夫链模型的运动估计新方法。实验结果表明,该方法能够对不同性质视频序列有很好的适应能力,并在计算成本和图像重建质量上得到了很好的折中。

Motion estimation is one of the most important parts of the video encoding/decoding system. Techniques of starting point prediction, stopping if good enough, as well as the different searching pattern are always adopted by fast block-matching motion estimation algorithms to improve their efficiency. A new method based on Markov Chain Model is proposed. By inducting the MarkowChain model, accuracy prediction of the starting point is achieved. And, by exploiting genetic algorithm and fuzzy logic, it can efficiently avoid being trapped in local optimum. Experimental results illustrate that the proposed method can attain good performance among different video sequences with different motion activity.