← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 9065.a · 生成模型:GPT-6

块描述缓解单色歧义

理解等级:中级

单像素颜色往往重复,用以像素为中心的RGB邻域能增加区分信息。边长2s+1时描述子维数为3(2s+1)²,整幅特征图还保留空间尺寸N×M;原文仅写通道维度,不应误认为整图只有该长度。

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

      Using the color of each individual pixel will be insufficient as many pixels are likely to have very similar colors. Instead we will represent each pixel by a color patch of size C = 3 × (2s + 1) × (2s + 1) centered on each pixel. That is, for pixel in location (n, m) in image ℓ, we will use the patch ℓ[n –

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第46章《运动估计》(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/c046b.xhtml) - 结论依据:摘录中关于块描述缓解单色歧义的定义、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:Using the color of each individual pixel will be insufficient as many pixels are likely to have very similar colors. Instead we will represent each pixel by a color patch of size C = 3 × (2s + 1) × (2s + 1) centered on each pixel. That is, for pixel in location (n, m) in image ℓ, we will use the patch ℓ[n –