← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8911 · 生成模型:GPT-6

加窗在成本和波纹间取舍

理解等级:中级

把理想低通核乘有限窗可缩短支撑。空间相乘使频谱与窗口频谱卷积,因此矩形硬截止变成有过渡带和波纹的响应。矩形、Hamming等窗口控制主瓣宽度与旁瓣衰减的取舍,不能同时无成本获得短核、陡截止和零波纹。

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

      One disadvantage of the ideal anti-aliasing filter is that the it requires convolving the image with a very large filter (as large as the image). We would like to use smaller filters to reduce the computational cost. One method to design anti-aliasing filters with a finite impulse response is to use a window, w[n], and to use as anti-aliasing filter the kernel w[n]h[n], where h[n]

    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:摘录中关于加窗在成本和波纹间取舍的定义、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:One disadvantage of the ideal anti-aliasing filter is that the it requires convolving the image with a very large filter (as large as the image). We would like to use smaller filters to reduce the computational cost. One method to design anti-aliasing filters with a finite impulse response is to use a window, w[n], and to use as anti-aliasing filter the kernel w[n]h[n], where h[n]