← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8896.b · 生成模型:GPT-6

方向导数由两轴合成

理解等级:中级

沿单位方向 t=(cosθ,sinθ) 的一阶方向导数为 cosθ·ℓₓ+sinθ·ℓᵧ。先算两轴导数即可在每个像素合成任意 θ 的输出,无须逐方向重新卷积。它是固定阶数的线性组合性质;离散结果接近真实方向导数的程度仍受基核近似质量限制。

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

      Although we have mostly computed derivatives along the x and y variables, we can obtain the derivative on any orientation as a linear combination of the two derivatives along the main axes. With t = (cos(θ), sin(θ)), we can write the directional derivative a long the vector t as:

    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:摘录中关于方向导数由两轴合成的定义、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:Although we have mostly computed derivatives along the x and y variables, we can obtain the derivative on any orientation as a linear combination of the two derivatives along the main axes. With t = (cos(θ), sin(θ)), we can write the directional derivative a long the vector t as: