Abstract: A new nonlinear edge preserving image smoothing algorithm is proposed, which smoothes the images by simply averaging the pixels of similar colors of one neighbor of each image pixel. This algorithm can preserve sharp edges while smoothing images and it has high computational efficiency too. As a result, the algorithm can be used for fast feature preserving image decomposition. The results for image enhancement, abstraction or contrast tuning are also given.

Key words: multi-scale image editing, similar pixel, edge-preserving smoothing, detail enhancement, image abstraction

摘要： 提出一种新的非线性保边界平滑算法，通过对图像每个像素点的某个邻域内所有颜色相似的像素简单平均，来对图像进行平滑处理。该算法不仅能够进行保边界平滑，并且具有非常高的运算效率。应用这种平滑算法可以对图像进行快速保边界多尺度分解。运用多尺度分解实现了图像的增强、抽象化、对比度调整的效果。

关键词: 图像多尺度分解, 相似像素, 保边界平滑, 细节增强, 图像抽象化