知识卡片

普通读书笔记卡 · 9077.b · 生成模型:GPT-6

图像金字塔缩小表观位移

理解等级:中级

在低分辨率层,同一运动对应较少像素位移,更适合一阶近似。先低通再降采样可减轻混叠,但细小物体与边界也可能消失,所以粗到细并不保证解决所有大运动或薄结构匹配。

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

      One advantage of this approach over the matching-based algorithm is that the estimated flow is not discrete, but it does not work well if the displacement is large, as the Taylor approximation is not valid anymore. One solution to the problem of large displacements is to compute a Gaussian pyramid for the input sequence. The low-resolution scales make the motion smaller and the gradient-based approach

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第48章《光流估计》(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/c048.xhtml) - 结论依据:摘录中关于图像金字塔缩小表观位移的定义、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:One advantage of this approach over the matching-based algorithm is that the estimated flow is not discrete, but it does not work well if the displacement is large, as the Taylor approximation is not valid anymore. One solution to the problem of large displacements is to compute a Gaussian pyramid for the input sequence. The low-resolution scales make the motion smaller and the gradient-based approach