← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8890.b · 生成模型:GPT-6

高斯复合时相加的是方差

理解等级:中级

连续归一化高斯卷积仍为高斯,σ结果²=σ₁²+σ₂²。重复两次σ滤波相当于一次√2σ,而不是2σ;若从已有模糊补到更大目标尺度,可据方差差确定新增连续核。离散近似和边界会带来偏差。

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

      the variance of the result is the sum . This is a remarkable property of Gaussian filters and is the basis of the Gaussian pyramid that we will see later in chapter 23. To prove this property, one can use the Fourier transform of the Gaussian and the fact that the convolution is the product of Fourier transforms.

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第17章《模糊滤波器》(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/c017b.xhtml) - 结论依据:摘录中关于高斯复合时相加的是方差的定义、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:the variance of the result is the sum . This is a remarkable property of Gaussian filters and is the basis of the Gaussian pyramid that we will see later in chapter 23. To prove this property, one can use the Fourier transform of the Gaussian and the fact that the convolution is the product of Fourier transforms.