← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 9078.d · 生成模型:GPT-6

分层模型联合分配与运动

理解等级:中级

将场景近似为若干刚性运动层,可交替估计像素归属及每层参数,采用类似EM的组织方式。层次遮挡提供额外结构,但单个刚体一般仍有逐点深度变化,不能无条件简化为全层同一二维平移或单应;层数、初始化及分配歧义也需处理。

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

      In this case, instead of a moving camera we have rigid moving objects (which is equivalent). The 2D motion field can then be represented as a collection of superimposed layers, each layer containing one object and occluding the layers below. Each layer will be described by a different set of motion parameters. The parametric motion field can be incorporated into the gradient-based approach described previously.

    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:摘录中关于分层模型联合分配与运动的定义、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:In this case, instead of a moving camera we have rigid moving objects (which is equivalent). The 2D motion field can then be represented as a collection of superimposed layers, each layer containing one object and occluding the layers below. Each layer will be described by a different set of motion parameters. The parametric motion field can be incorporated into the gradient-based approach described previously.