← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8959 · 生成模型:GPT-6

邻域匹配形成条件采样

理解等级:中级

Efros–Leung方法在参考图中搜索与已生成邻域相似的位置,从合格匹配位置的中心像素随机选值,填入当前未知像素。已生成像素提供条件,参考图提供经验候选;它不是把多个中心值直接平均,因此可避免平均产生不存在的中间颜色。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      At each iteration, we have synthesized pixels in the neighborhood of the pixel to be synthesized. We look through the input image for examples of similar configurations of pixel values as the current local neighborhood, having a sum of squared intensity differences from the local neighborhood lower than some threshold. Randomly select one of the matching instances, and copy the value of the pixel into

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第28章《纹理》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c028.xhtml) - 结论依据:摘录中关于邻域匹配形成条件采样的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:At each iteration, we have synthesized pixels in the neighborhood of the pixel to be synthesized. We look through the input image for examples of similar configurations of pixel values as the current local neighborhood, having a sum of squared intensity differences from the local neighborhood lower than some threshold. Randomly select one of the matching instances, and copy the value of the pixel into