← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8924.c · 生成模型:GPT-6

保留最粗残差才能重建

理解等级:中级

保留全部lₖ和最粗低通层gₙ后,按从粗到细递推gₖ=lₖ+Fₖgₖ₊₁,可恢复g₀。若丢弃最粗层,整体低频结构通常无法恢复;只保存“边缘细节”不构成完整表示。重建应使用编码时相同的Fₖ、边界和尺寸对齐。

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

      Interestingly, the Laplacian pyramid is an invertible transform, but only if we keep the low-pass residual. We can reconstruct the original image from the Laplacian pyramid. Using the low-pass residual signal associated with the Laplacian pyramid, we can recursively reconstruct the corresponding Gaussian pyramid. Remember that in the Gaussian pyramid level 1 is just the original image itself, so we can use this to reconstruct

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第23章《图像金字塔》(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/c023.xhtml) - 结论依据:摘录中关于保留最粗残差才能重建的定义、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:Interestingly, the Laplacian pyramid is an invertible transform, but only if we keep the low-pass residual. We can reconstruct the original image from the Laplacian pyramid. Using the low-pass residual signal associated with the Laplacian pyramid, we can recursively reconstruct the corresponding Gaussian pyramid. Remember that in the Gaussian pyramid level 1 is just the original image itself, so we can use this to reconstruct