← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8885.a · 生成模型:GPT-6

循环位移只改变频谱相位

理解等级:中级

将长度N序列循环位移n₀,频谱乘exp(−2πjun₀/N),模保持不变。真实相机平移常引入新的边界像素和遮挡,因而不满足精确循环位移假设;幅度近似不变可作线索,却不能当所有视角变化的严格不变性。

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

      this property is only true if we apply a circular shift, i.e., pixels that disappear on the boundary due to the translation reappear on the other side. This is a consequence of the periodic structure of the signal ℓ assumed by the DFT. In practice, when the translation is due to the motion of the camera, the shift will not be circular and new pixels

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第16章《傅里叶分析》(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/c016.xhtml) - 结论依据:摘录中关于循环位移只改变频谱相位的定义、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:this property is only true if we apply a circular shift, i.e., pixels that disappear on the boundary due to the translation reappear on the other side. This is a consequence of the periodic structure of the signal ℓ assumed by the DFT. In practice, when the translation is due to the motion of the camera, the shift will not be circular and new pixels