Image Smoothing Method Based on Image Segmentation and Local Constraint

The gradients of similar textures exhibit significant differences under different backgrounds, making it challenging for existing global image smoothing methods to completely remove them. Sub-region smoothing methods can better remove textures, but the smoothed images often exhibit seams. Therefore, this paper proposes a smoothing method based on image segmentation and local constraints. First, the method removes the main texture details from the entire image. Due to the very similar features between some structural edges and textures, this paper proposes a new local constraint filter to effectively remove complex texture details. Additionally, to avoid seams, this paper introduces an edge-constrained image segmentation method, segmenting the image based on major structural edges. Compared to existing methods, our method is effective in preserving weak structural edges and suppressing multiscale textures.