知识卡片

普通读书笔记卡 · 8895.b · 生成模型:GPT-6

混合导数保持可分离

理解等级:中级

各向同性二维高斯可分离成两个一维高斯,其 x 的 n 阶、y 的 m 阶混合导数也可分离为对应一维导数的乘积。因此可以按两个轴先后滤波,降低计算量。可分离性是笛卡尔方向的计算性质,并不意味着每个高阶核都旋转不变。

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

      In two dimensions, as the Gaussian is separable, the partial derivatives result on the product of two Hermite polynomial, one for each spatial dimension: (18.20) Figure 18.10 shows the 2D Gaussian derivatives, and figure 18.11 shows the corresponding Fourier transforms. Figure 18.12 shows the output of the derivatives when applied to a simple input image containing a square and a circle. Different derivatives detect a

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第18章《图像导数》(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/c018.xhtml) - 结论依据:摘录中关于混合导数保持可分离的定义、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:In two dimensions, as the Gaussian is separable, the partial derivatives result on the product of two Hermite polynomial, one for each spatial dimension: (18.20) Figure 18.10 shows the 2D Gaussian derivatives, and figure 18.11 shows the corresponding Fourier transforms. Figure 18.12 shows the output of the derivatives when applied to a simple input image containing a square and a circle. Different derivatives detect a