知识卡片

普通读书笔记卡 · 8952.d · 生成模型:GPT-6

高斯先验导出维纳收缩

理解等级:高级

零均值高斯先验协方差为C、独立白高斯噪声方差为σ²时,后验均值与MAP一致,为C(C+σ²I)⁻¹y。频域每个分量乘Sx/(Sx+σ²):信号相对噪声越弱,收缩越强。非零均值需先减均值再加回;不同噪声协方差需要相应调整。

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

      The solution to this problem can be obtained in closed form: (27.11) This is just a linear operation. It can also be written in the Fourier domain as: (27.12) Figure 27.14 shows the result of using the Gaussian image model for image denoising. Although there is a certain amount of noise reduction, the result is far from satisfactory. The Gaussian image model fails to capture

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第27章《统计图像模型》(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/c027b.xhtml) - 结论依据:摘录中关于高斯先验导出维纳收缩的定义、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:The solution to this problem can be obtained in closed form: (27.11) This is just a linear operation. It can also be written in the Fourier domain as: (27.12) Figure 27.14 shows the result of using the Gaussian image model for image denoising. Although there is a certain amount of noise reduction, the result is far from satisfactory. The Gaussian image model fails to capture