基于全变差和脊波变换的海岸线提取算法

为有效解决视频图像中海岸线的自动提取问题,提出一种结合全变差增强水陆边缘和脊波变换预测搜索区域的海岸线提取算法.首先基于最小化全变差模型降噪和增强边缘;第2步,根据水陆像素的彩色特征,计算彩色梯度向量;第3步,利用脊波变换预测水陆交界处的大致方向和动态范围,缩小检测区域;最后用一种改进的最小代价梯度跟踪模型对海岸线进行搜索,跟踪并精确提取最终的海岸线定位参数.实验结果证明了算法的有效性和较强的抗噪声能力,并且对不同光照条件具有鲁棒性,是视频图像中提取海岸线的新方法,为下一步设计应用系统打下了基础.

Shoreline detection algorithm based on total variation and ridgelet transform

To detect shoreline from nearshore video images, a computational algorithm based on total variation minimization and ridgelet transform is proposed. Initially, total variation minimization model is applied to carry out image denoising and edge preserving. Secondly, the gradient vector in color space is calculated based on the color characteristics of land and water imagery and used to present waterline edges efficiently. Thirdly, the ap-proximate direction and area of coastline is obtained using ridgelet transform based linear feature extraction. Then the predicted image region of land/water boundaries is obtained using inverse Radon transform. Furthermore, the shoreline is determined by a searching approach of minimizing cost function. Experimental results indicate that the proposed method is effective and valid in finding the shoreline and is robust for noisy images under various lighting conditions. The proposed algorithm provides a basis for designing application systems afterwards.