知识卡片

普通读书笔记卡 · 8895 · 生成模型:GPT-6

导数尺度改变所见结构

理解等级:中级

小 σ 的高斯导数保留细纹理变化,大 σ 抑制细节后强调更大范围的结构。尺度选择改变“变化”的空间范围,也改变响应幅度;要跨尺度比较强弱,需要另外规定尺度归一化,不能把不同 σ 的原始输出直接当成等价证据。

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

      Figure 18.8 shows an image filtered with three Gaussian derivatives with different widths, σ. Derivatives at different scales emphasize different aspects of the image. The fine-scale derivatives (figure 18.8[a]) highlight the bands in the zebra, while the coarse-scale derivatives (figure 18.8[c]) emphasize more the object boundaries. This multiscale image analysis will be studied in depth in the following chapter.

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第18章《图像导数》(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/c018.xhtml) - 结论依据:摘录中关于导数尺度改变所见结构的定义、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:Figure 18.8 shows an image filtered with three Gaussian derivatives with different widths, σ. Derivatives at different scales emphasize different aspects of the image. The fine-scale derivatives (figure 18.8[a]) highlight the bands in the zebra, while the coarse-scale derivatives (figure 18.8[c]) emphasize more the object boundaries. This multiscale image analysis will be studied in depth in the following chapter.