知识卡片

普通读书笔记卡 · 8879.a · 生成模型:GPT-6

归一化相关比较中心化方向

理解等级:高级

标准零均值归一化相关将模板h与图块p分别中心化,再除以各自L2范数后做内积。非零方差时它对图块的加性偏移及正比例缩放不敏感,值在[−1,1]。常量模板或图块分母为零需单独处理,不能直接输出可靠相似度。

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

      normalized to be zero mean and unit norm) and the local image patch, p, centered at that location and with the same size as the kernel, normalized to be unit norm. This can be implemented with the following steps: First, we normalize the kernel h. The normalized kernel is the result of making the kernel h zero mean and unit norm. Then, the normalized cross-correlation

    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:摘录中关于归一化相关比较中心化方向的定义、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:normalized to be zero mean and unit norm) and the local image patch, p, centered at that location and with the same size as the kernel, normalized to be unit norm. This can be implemented with the following steps: First, we normalize the kernel h. The normalized kernel is the result of making the kernel h zero mean and unit norm. Then, the normalized cross-correlation