目的 颜色编辑扩散是图像处理领域中的一个重要研究内容。以往方法中,为了取得期望的编辑效果,要求交互选定的样本颜色尽可能丰富。为了降低输入条件,提出一种新的有效重采样的颜色编辑扩散技术。方法 首先对图像做超像素分割并提取每个超像素中心作为重采样样本。遵循超像素中心非边缘像素,只服从一种编辑模式的原则,计算其与每种编辑模式的相似度,如果与某种编辑模式相似度绝对占优,则该样本的编辑模式直接确定。对于待定样本,依照相邻相似样本服从相同编辑模式的思想来判断其归属。然后将重采样样本的颜色编辑遵循相似颜色编辑后仍相似的原则采用增量的方式扩散到整幅图像。扩散时每个像素自适应地选择样本的数量,以避免相似度低样本的干扰。结果 与以往的方法进行比较,该方法对输入条件的敏感度较低,且在相同的相对简单的输入条件下能取得更好的视觉效果,可视化的梯度显示细节保持也更理想。结论 该方法能有效简化用户交互,将稀疏的画笔像素的扩散转化为颜色信息相对丰富的超像素中心样本的扩散,且生成的图像能忠于编辑颜色及较好地保持图像细节。

Objective Edit propagation is an important operation in image processing. Previous methods require that the samples selected interactively should have a sufficiently large number of colors. In this study, we propose a new robust edit propagation technique by efficient resampling, which may reduce the requirement of interactive operation. Method First, the image is segmented into superpixels and the center pixel of each superpixel is considered a new sample. Following the rule that each sample obeys one kind of editing mode, we calculate the edited color of each new sample by the similarity between the sample and editing modes and the similarity between those samples and its neighbors. The sample whose similarity to several editing modes is evidently high directly belongs to the corresponding editing mode. Otherwise, the editing mode of the current sample is determined by the neighboring samples with high similarities to the current sample. Then, the edited colors of all samples are propagated into the entire image based on the policy that similar colors are still similar after editing. Based on the similarity of the pixel and the samples, we select the number of samples adaptively for each pixel to reduce the interference of those samples with low similarity. Result Compared with previous methods, our solution is less sensitive to the input strokes. With the same simple strokes, our results exhibit better vision and details. Conclusion The method can effectively simplify the user interaction by replacing the sparse input samples with many colors resamples. Moreover, the experiments and comparisons show that the method can preserve the edited colors and its details.