Structure-preserving image smoothing is used in solving many problems of video information processing and analysis. Most smoothing methods are based on indirect information about the characteristics of a local image area, for example, the distribution of signal values, rather than on direct data of the image spatial structure. In reality, a criterion for constraining the smoothing area should not be the distribution of signal values or the difference in brightness of the objective and surrounding points, but their belonging to the same spatial area of an image. A criterion sufficient for the connectivity of points in an area is the absence of contour lines between them. An approach to image smoothing is proposed, which is based on preliminary detection of contour edges between image areas and subsequent contour-constrained smoothing inside each area. A concept of “affinity” of image points is introduced, on the basis of which the smoothing algorithm is built. The proposed algorithm is experimentally compared with other available smoothing algorithms.

保结构图像平滑用于解决视频信息处理和分析的许多问题。大多数平滑方法是基于关于局部图像区域特征的间接信息，例如信号值的分布，而不是基于图像空间结构的直接数据。实际上，约束平滑区域的标准不应该是信号值的分布或目标和周围点的亮度差异，而应该是它们属于图像的同一空间区域。一个区域中点的连通性足够的标准是它们之间没有等高线。提出了一种图像平滑的方法，它基于图像区域之间轮廓边缘的初步检测和每个区域内的后续轮廓约束平滑。引入了图像点“亲和性”的概念，在此基础上构建了平滑算法。所提出的算法与其他可用的平滑算法进行了实验比较。