← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8844.b · 生成模型:GPT-6

正则化逆需按实际公式计算

理解等级:高级

本章正则解为ℓw=Bℓs,B=(AᵀA+λI)⁻¹Aᵀ。若A=I,则B=I/(1+λ),λ非零时不是I,原图注把三者都说成单位矩阵不准确。正则化也不普遍保证图像更真实,应说明先验及强度。

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

      Where the matrix B = (AT A + λI)–1AT is the regularized inverse of the imaging matrix A. 1D and the scene lives in flatland. Let the sensor measurements be ℓS and the unknown scene be ℓw. We represent the camera by matrix A. For the case of a pinhole camera, and assuming 13 pixel sensor observations, the camera matrix is just a 13×13 identity

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第07章《作为线性系统的相机》(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/c007.xhtml) - 结论依据:摘录中关于正则化逆需按实际公式计算的定义、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:Where the matrix B = (AT A + λI)–1AT is the regularized inverse of the imaging matrix A. 1D and the scene lives in flatland. Let the sensor measurements be ℓS and the unknown scene be ℓw. We represent the camera by matrix A. For the case of a pinhole camera, and assuming 13 pixel sensor observations, the camera matrix is just a 13×13 identity