live wire算法中局部代价计算的加速改进

目的加快livewire算法中局部代价权值的计算速度，提升算法的运行效率。方法提出利用计算相关性同步计算相邻节点的梯度方向权值代价分量^，以及只计算最短路径搜索中已排序节点相对于邻域节点的权值代价两种方法改进局部代价算法。结果改进算法对于不同图像空间分辨率的情况，代价权值的计算量，livewire算法的运行时间以及交互分割时间上，相比较其他两种算法都有一定程度的减少。结论改进代价计算方式的livewire算法，能够正确分割目标图像，在高图像分辨率的情况下明显提高了交互分割速度。

Acceleration of local cost computation in live wire algorithm

Objective To accelerate local cost computation in live wire algorithm. Methods Two methods were proposed to improve the efficiency of local cost computation including synchronously calculating fD local cost components between the neighboring pixels using correlation information and only calculating local costs of the processed pixels in the optimal path searching. Results Compared with two other algorithms, the improved algorithm had less workload of local cost computation,less running time and interactive segmentation time for different image resolutions. Conclusions Improved algorithm of local cost computation accelerated the speed of interactive segmentation of live wire algorithm especially in high image resolution.