知识卡片

普通读书笔记卡 · 9030.c · 生成模型:GPT-6

半全局匹配是多路径近似

理解等级:中级

SGM沿多个方向进行一维路径代价聚合,以较低代价近似全局立体匹配目标。多个路径并不把二维带环优化自动变成精确树推断;路径间重复与遮挡、边界处理仍影响结果。其意义是计算与空间一致性的折中。

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

      Methods to interpolate depth between the depths of matched points include probabilistic methods such as belief propagation, discussed in chapter 29. Under assumptions about the local spatial relationships between depth values, optimal shapes can be inferred provided the spatial structure over which inference is performed is a chain or a tree. Note that [207] uses a related inference algorithm, dynamic programming, which has the same

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第40章《立体视觉》(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/c040.xhtml) - 结论依据:摘录中关于半全局匹配是多路径近似的定义、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:Methods to interpolate depth between the depths of matched points include probabilistic methods such as belief propagation, discussed in chapter 29. Under assumptions about the local spatial relationships between depth values, optimal shapes can be inferred provided the spatial structure over which inference is performed is a chain or a tree. Note that [207] uses a related inference algorithm, dynamic programming, which has the same