← 返回刚才阅读的位置

知识卡片

结构图卡 · 8913.c · 生成模型:GPT-6

扩展与插值分两步

理解等级:中级

结构图:

flowchart LR
 A[低分辨率样本] --> B[按整数倍率插零扩展]
 B --> C[应用插值核]
 C --> D[高分辨率表示]
 E[核的支撑与增益] --> C
 F[坐标对齐及边界] --> B
 F --> C

扩展只改变样本网格,插值才填补间隔;两步都不提供额外场景观测。

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

      Upsampling consists in the sequence of expansion followed by interpolation: (21.31) (21.32) Upsampling by a factor of 2 is done by inserting one 0 between each two samples (first for the horizontal dimension and then for the vertical one), and then interpolating between samples applying the binomial filter [1/2, 1, 1/2] ◦ [1/2, 1, 1/2]T. Figure 21.21 shows the upsampling process for a color image.

    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:摘录中关于扩展与插值分两步的定义、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:Upsampling consists in the sequence of expansion followed by interpolation: (21.31) (21.32) Upsampling by a factor of 2 is done by inserting one 0 between each two samples (first for the horizontal dimension and then for the vertical one), and then interpolating between samples applying the binomial filter [1/2, 1, 1/2] ◦ [1/2, 1, 1/2]T. Figure 21.21 shows the upsampling process for a color image.