← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8976.c · 生成模型:GPT-6

像素特征决定聚类含义

理解等级:中级

将每个像素变成一行特征,就可用聚类获得分割标签。若只用Lab中的a、b色度,分组主要由色彩差异决定,并忽略亮度L。该选择可让黑白条纹在色度空间接近,却可能把不同位置、不同物体的相近色像素混为一组。

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

      K-means operates on a data matrix of N datapoints by M features per datapoint. Our first task is to convert the segmentation problem to that format. The simplest thing to do is let each pixel be a datapoint and the pixel's color be its feature vector. For ease of visualization, we will use the ab color value as the two-dimensional (2D) feature per pixel. An

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第31章《知觉分组》(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/c031.xhtml) - 结论依据:摘录中关于像素特征决定聚类含义的定义、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:K-means operates on a data matrix of N datapoints by M features per datapoint. Our first task is to convert the segmentation problem to that format. The simplest thing to do is let each pixel be a datapoint and the pixel's color be its feature vector. For ease of visualization, we will use the ab color value as the two-dimensional (2D) feature per pixel. An