← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8950 · 生成模型:GPT-6

直方图要归一成概率

理解等级:中级

灰度直方图记录各强度出现次数,除以总像素数后才得到概率;RGB联合直方图可用256³个离散颜色格。多图汇总需明确按像素或按图平均,不同尺寸下二者权重不同。独立抽样匹配的是分布,单次有限样本不必精确复制训练直方图。

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 parameters, θ, of the distribution are those 256 (or 2563 for color) values. If we are given a set of images, we can estimate the vector θ as the average histogram of the images. We can also estimate the model from a single image. We can then sample new images based on equation (27.1). In order to sample a new image, we will sample

    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:摘录中关于直方图要归一成概率的定义、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 parameters, θ, of the distribution are those 256 (or 2563 for color) values. If we are given a set of images, we can estimate the vector θ as the average histogram of the images. We can also estimate the model from a single image. We can then sample new images based on equation (27.1). In order to sample a new image, we will sample