← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8877 · 生成模型:GPT-6

均值核保直流但会受边界影响

理解等级:中级

5×5核各项为1/25,在内部计算局部均值,抑制快速变化并使细节模糊。核权重和为1可保持常量输入,但零填充边界会掺入虚构暗值,导致边缘变暗。不能仅由归一化核断言有限图像所有位置亮度都保持。

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

      a width of two pixels. Figure 15.10(c) shows the result of convolving the image with a kernel 0.5δ [n – 2, m – 2] + 0.5δ [n + 2, m + 2]. This results in two superimposed copies of the same image shifted diagonally in opposite directions. And finally, figure 15.10(d) shows the result of convolving the image with a uniform kernel of all 1/25.

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第15章《线性图像滤波》(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/c015b.xhtml) - 结论依据:摘录中关于均值核保直流但会受边界影响的定义、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:a width of two pixels. Figure 15.10(c) shows the result of convolving the image with a kernel 0.5δ [n – 2, m – 2] + 0.5δ [n + 2, m + 2]. This results in two superimposed copies of the same image shifted diagonally in opposite directions. And finally, figure 15.10(d) shows the result of convolving the image with a uniform kernel of all 1/25.