← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 9019.a · 生成模型:GPT-6

逆向采样逐个覆盖目标格

理解等级:中级

逆向映射遍历每个目标像素 q,用 p=M^−1 q 找源位置,再插值取颜色。只要变换可逆且源位置有效,就避免简单前向投点留下的内部空洞。超出源图、奇异变换和投影无穷远仍需明确边界处理;覆盖目标格不等于消除了混叠。

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

      The best approach is to use what is called backward mapping which consists of looping over all the pixels of the target image and applying the inverse geometric transform, M–1; we then use interpolation (as described in section 21.4.1 in chapter 21) to get the correct color value (figure 38.10). This process guarantees that there will be no missing values (unless the coordinates go outside

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第38章《图像与几何表示》(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/c038b.xhtml) - 结论依据:摘录中关于逆向采样逐个覆盖目标格的定义、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:The best approach is to use what is called backward mapping which consists of looping over all the pixels of the target image and applying the inverse geometric transform, M–1; we then use interpolation (as described in section 21.4.1 in chapter 21) to get the correct color value (figure 38.10). This process guarantees that there will be no missing values (unless the coordinates go outside