← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8891.d · 生成模型:GPT-6

最高频归零不等于理想抗混叠

理解等级:高级

正阶二项核可精确取消交替序列的π频率,但降采样2倍需关注整个新奈奎斯特界π/2以上频带,而非只看π点。简单核可减少混叠,却不能保证任意输入都无混叠。零阶冲激核也不具该消除作用。

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

      convolution of g5 [n] with the wave [1, –1, 1, –1, …] is not zero. This is to be expected from the form of the FT of the Gaussian. This is not a strong limitation and in many applications it does not matter. However, in some cases is important to completely cancel the highest frequencies (like when applying antialiasing filters as we will see later).

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第17章《模糊滤波器》(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/c017b.xhtml) - 结论依据:摘录中关于最高频归零不等于理想抗混叠的定义、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:convolution of g5 [n] with the wave [1, –1, 1, –1, …] is not zero. This is to be expected from the form of the FT of the Gaussian. This is not a strong limitation and in many applications it does not matter. However, in some cases is important to completely cancel the highest frequencies (like when applying antialiasing filters as we will see later).