← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8910.c · 生成模型:GPT-6

截止频率随倍率下降

理解等级:中级

抽取 k 倍后,每轴允许的归一化角频率上限缩为 π/k。居中的有限DFT索引则对应 |u|/(2k)、|v|/(2k) 的保留区域,具体端点需按输出奈奎斯特分量处理。缩小更多时不能沿用同一弱低通,否则剩余高频仍会混叠。

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

      Anti-aliansing filtering is an essential operation whenever we manipulate the resolution of an image. When decimating a signal of size N × N with a scaling factor k, aliasing is produced by all the content at frequencies that are above u, v > N/2k. Therefore, to reduce aliasing, we have to remove any spectral content of the signal present at high spatial frequencies. We will

    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:摘录中关于截止频率随倍率下降的定义、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:Anti-aliansing filtering is an essential operation whenever we manipulate the resolution of an image. When decimating a signal of size N × N with a scaling factor k, aliasing is produced by all the content at frequencies that are above u, v > N/2k. Therefore, to reduce aliasing, we have to remove any spectral content of the signal present at high spatial frequencies. We will