知识卡片
卷积的位移与支撑可推导
把输入或核平移同样位移,完整卷积结果随之平移。长度N与M的有限支撑序列,完整线性卷积最多占N+M−1个位置;具体非零范围还可能因零值或抵消缩小。库的same或valid输出是额外裁剪约定,不由支撑定理唯一决定。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 线性图像滤波
原书出处 · 1 处
- · 查看原书依据
Translating the kernel is equivalent: (15.19) Given the signal: The signal ℓ [n – n0] is a shifted version of ℓ [n]. With n0 = 2 this is Support. The convolution of a discrete signal of length N with another discrete signal of length M results in a discrete signal with length L ≤ M + N – 1. Identity. The convolution also has an
参考来源
- 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第15章《线性图像滤波》(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/c015b.xhtml)
- 结论依据:摘录中关于卷积的位移与支撑可推导的定义、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:Translating the kernel is equivalent: (15.19) Given the signal: The signal ℓ [n – n0] is a shifted version of ℓ [n]. With n0 = 2 this is Support. The convolution of a discrete signal of length N with another discrete signal of length M results in a discrete signal with length L ≤ M + N – 1. Identity. The convolution also has an