知识卡片

普通读书笔记卡 · 8909.c · 生成模型:GPT-6

抽取是降维线性映射

理解等级:中级

抽取矩阵每行仅选择一个输入位置,因此对输入值线性,但丢弃了大量自由度,通常无法逆转。相同输出可以来自不同的被丢弃像素。线性也不等于对任意单像素平移等变:输入平移会改变保留下来的网格相位。

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

      Decimation is a linear operation therefore we can write it as a matrix. In the one-dimensional (1D) case, for a signal of length 8, subsampling by a factor of 2 can be described by the following linear operator:

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第21章《图像下采样与上采样》(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/c021.xhtml) - 结论依据:摘录中关于抽取是降维线性映射的定义、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:Decimation is a linear operation therefore we can write it as a matrix. In the one-dimensional (1D) case, for a signal of length 8, subsampling by a factor of 2 can be described by the following linear operator: