知识卡片

普通读书笔记卡 · 8954.a · 生成模型:GPT-6

先验最优不等于典型图像

理解等级:中级

若带通响应的密度都在零处最大,常数图像可使所有响应为零,从而获得很高的先验值。它不代表自然图像通常是常数,也不意味着应忽略观测。恢复任务必须用观测约束与先验共同决定答案,众数和高维分布的典型样本也需区分。

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

      What is the most likely image under the model equation (27.14)? It is a constant image. To see how this model works, let's look at a toy problem. 27.6.1 Toy Model of Image Inpainting To build intuition about image structures that are likely under the wavelet marginal model, let's consider a simple 1D image of length N = 11 that has a missing value in

    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:摘录中关于先验最优不等于典型图像的定义、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:What is the most likely image under the model equation (27.14)? It is a constant image. To see how this model works, let's look at a toy problem. 27.6.1 Toy Model of Image Inpainting To build intuition about image structures that are likely under the wavelet marginal model, let's consider a simple 1D image of length N = 11 that has a missing value in