知识卡片

普通读书笔记卡 · 8893.a · 生成模型:GPT-6

边界决定差分可逆性

理解等级:中级

书中方阵差分矩阵保留第一个像素相对零边界的差,因此首值实际上提供了绝对亮度锚点;其逆矩阵通过累加差分恢复整条信号。所谓“从导数恢复图像”是否唯一,首先取决于边界是否已经给出常数分量,而不只取决于用了导数表示。

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 simple way of seeing that we can recover the input from its derivatives is to write the derivative in matrix form. This is the matrix that corresponds to the convolution with the kernel [1, –1] that we will call D0. The next two matrices show the matrix D0 and its inverse D0–1 for a 1D image of length five pixels using zero boundary conditions:

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第18章《图像导数》(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/c018.xhtml) - 结论依据:摘录中关于边界决定差分可逆性的定义、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 simple way of seeing that we can recover the input from its derivatives is to write the derivative in matrix form. This is the matrix that corresponds to the convolution with the kernel [1, –1] that we will call D0. The next two matrices show the matrix D0 and its inverse D0–1 for a 1D image of length five pixels using zero boundary conditions: