← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8958.c · 生成模型:GPT-6

参数纹理模型依赖选定统计

理解等级:中级

参数方法用固定统计向量概括参考图,存储可与原图尺寸解耦,也便于比较或改变参数。代价是未被统计记录的信息会丢失;维数多少取决于具体模型,书中约千维是实例量级,不是“参数纹理模型”的定义性常数。

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

      Parametric texture models represent a texture image with a small vector of parameters, usually around 1,000 dimensions [392]. These models, usually called parametric texture models, stayed popular until 2000. After that, non-parametric texture models became dominant (section 28.4) until generative models with neural networks emerged around 2014. 28.4 Efros-Leung Texture Analysis and Synthesis Model Nonparametric Markov random field image models, introduced earlier in chapter 27,

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第28章《纹理》(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/c028.xhtml) - 结论依据:摘录中关于参数纹理模型依赖选定统计的定义、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:Parametric texture models represent a texture image with a small vector of parameters, usually around 1,000 dimensions [392]. These models, usually called parametric texture models, stayed popular until 2000. After that, non-parametric texture models became dominant (section 28.4) until generative models with neural networks emerged around 2014. 28.4 Efros-Leung Texture Analysis and Synthesis Model Nonparametric Markov random field image models, introduced earlier in chapter 27,