知识卡片

普通读书笔记卡 · 8952.b · 生成模型:GPT-6

频谱相同仍可失去对象

理解等级:中级

随机化相位的图像可保留主方向与细节尺度,却通常不再显示原有物体布局。头发等强方向纹理可能仍较相似,说明适用性取决于内容。频谱描述能量分布,不能独自规定跨频率相位如何协同形成边缘和形状。

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

      As shown in figure 27.12, when fitting the Gaussian model to different real images, and then sampling from it, the result is always an image that looks like cloudy noise regardless of the image content. However, a few attributes of the original image are preserved including the size of the details present in the image and the dominant orientations.

    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:摘录中关于频谱相同仍可失去对象的定义、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:As shown in figure 27.12, when fitting the Gaussian model to different real images, and then sampling from it, the result is always an image that looks like cloudy noise regardless of the image content. However, a few attributes of the original image are preserved including the size of the details present in the image and the dominant orientations.