基于动态规划的最优化医学超声图像边缘提取

动态规划（DP）是一种解决多阶段决策过程最优化的方法。图像边缘提取时为了使系统输出具有最小的不确定性，考虑最优化判据是必要的。动态规划算法用于图像的边缘检测主要是获得一个图像的最低能量代价阵的过程，而图像的边缘对应于最低代价阵中的终止点和起始点之间能量梯度降低最快的路径，由此可以由最低代价阵勾勒出需要的边缘。对于质量较差的图像，可以先用梯度算子和一种LUM滤波器相结合进行预滤波。实验表明，基于该算法用于超声图像的边缘检测要获得全局最优的稳定的边缘线，是一种理想的最优化算法。

Dynamic Programming Based Optimal Edge Detection Algorithms on B-Scan Ultrasound Images

The segmentation of biomedical objects from gray scale images has been extensively studied in recent years. This paper explored a novel dynamic programming (DP) based optimal edge detection technique in medical ultrasound images. Dynamic programming is an optimal approach in multiscale decision-making. In an image segmentation system, it is used to obtain global optimal contours with connectedness and closeness. The DP algorithms process the object image to get the least comulative cost matrix to tracking a global optimal boundary. Combining with low-upper-middle (LUM) nonlinear enhancement filter and Gaussian preprocessor, this method shows robustness on the noisy images boundary detection.