知识卡片
整数块匹配缺少亚像素精度
只搜索整数偏移会把真实连续位移量化,尤其影响微小运动与精密测量。插值代价或对图像进行双线性、双三次采样可以支持细化,但仍需适当优化与外观假设;插值并不自动消除所有误差。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 运动估计
原书出处 · 1 处
- · 查看原书依据
This approach has many shortcomings. To start, we assumed that motion is discrete (it can only take on integer values). Therefore, the current approach does not compute displacements of pixels to subpixel accuracy that might be important if we need precision or if motions are very small. We could improve the approach by interpolating the cost function or by computing subpixel displacements using bilinear or
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:摘录中关于整数块匹配缺少亚像素精度的定义、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:This approach has many shortcomings. To start, we assumed that motion is discrete (it can only take on integer values). Therefore, the current approach does not compute displacements of pixels to subpixel accuracy that might be important if we need precision or if motions are very small. We could improve the approach by interpolating the cost function or by computing subpixel displacements using bilinear or