← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8953 · 生成模型:GPT-6

去噪残差可以揭示失真

理解等级:中级

将观测减去去噪结果,检查残差是否残留物体边缘和纹理,可以发现模型把真实结构当作噪声移除的问题。残差有结构是诊断线索,而非单独的完整质量标准;应同时检查输出、已知噪声假设及有参考时的恢复误差。

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

      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 the properties that make a natural image look real.

    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:摘录中关于去噪残差可以揭示失真的定义、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: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 the properties that make a natural image look real.