← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8898.c · 生成模型:GPT-6

雷蒂内克先验限制分解

理解等级:中级

本章 Retinex 假设照明平滑、反射率由近似均匀区域及尖锐边界组成。在对数域把乘积化为加法,再把较大的局部导数归给反射率、小导数归给照明。硬阴影、渐变材质和纹理会违背这种分工,因此阈值并非物理原因的可靠判别器。

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

      But how can we estimate r(x, y) and l(x, y) by only observing ℓ(x, y)? One of the first solutions to this problem was the Retinex algorithm proposed by Land and McCann [283]. Land and McCann observed that it is possible to extract r and l from ℓ if one takes into account some constraints about the expected nature of the images r and l.

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第18章《图像导数》(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/c018.xhtml) - 结论依据:摘录中关于雷蒂内克先验限制分解的定义、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:But how can we estimate r(x, y) and l(x, y) by only observing ℓ(x, y)? One of the first solutions to this problem was the Retinex algorithm proposed by Land and McCann [283]. Land and McCann observed that it is possible to extract r and l from ℓ if one takes into account some constraints about the expected nature of the images r and l.