知识卡片

普通读书笔记卡 · 9067.c · 生成模型:GPT-6

先估运动可为分组提供线索

理解等级:中级

块匹配不需预先知道对象类别或给定实例分割,因此所得运动可作为后续分组输入。它仍依赖局部外观相似、有限搜索和块内近似一致运动,并非完全没有场景假设,也不保证自动得到对象边界。

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 algorithm is that motion is computed in a way that is independent of the objects present in the scene. We did not make any assumption about what objects are moving or how. We did not introduce any grouping cues (as in chapter 31) to presegment the image into candidate objects. Therefore, we could use the computed motion as new cue for

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第46章《运动估计》(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/c046b.xhtml) - 结论依据:摘录中关于先估运动可为分组提供线索的定义、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 algorithm is that motion is computed in a way that is independent of the objects present in the scene. We did not make any assumption about what objects are moving or how. We did not introduce any grouping cues (as in chapter 31) to presegment the image into candidate objects. Therefore, we could use the computed motion as new cue for