知识卡片

普通读书笔记卡 · 8877.b · 生成模型:GPT-6

边界扩展表达未观测区域假设

理解等级:中级

有限图像外的真实像素不可见,边界扩展是对它们的假设:零填充假定暗背景,循环填充将另一侧接入,镜像延续反射结构,重复填充延续最近边值。各有伪影,某张图镜像误差较小不证明它对所有任务最优。

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

      Boundary extensions are different ways of approximating the ground truth image that exists beyond the image boundary. The top row shows how the image is extended beyond its boundary when using different types of boundary extensions (only the central part corresponds to the input). The middle row shows the output of the convolution (cropped to match the size of the input image). We can see

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第15章《线性图像滤波》(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/c015b.xhtml) - 结论依据:摘录中关于边界扩展表达未观测区域假设的定义、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:Boundary extensions are different ways of approximating the ground truth image that exists beyond the image boundary. The top row shows how the image is extended beyond its boundary when using different types of boundary extensions (only the central part corresponds to the input). The middle row shows the output of the convolution (cropped to match the size of the input image). We can see