图像处理Chap 5 Geometric Operations.ppt

Chap 5 Geometric Operations Purpose change spatial relationship among pixels in an image. Chap 5 Geometric Operations Two Considerations Grey-level interpolation (插值) Spatial transformation Chap 5 Geometric Operations 点运算 对单幅图像做处理，不改变像素的空间位置； 代数运算 对多幅图像做处理，也不改变像素的空间位置； 几何运算 对单幅图像做处理，改变像素的空间位置； 几何运算的两个独立算法： 灰度级插值算法 (Grey-level Interpolation) 空间变换算法 (Spatial Transformation) Chap 5 Geometric Operations Consider: 设: 输入图像大小为(w, h)，输出图像大小为(W, H) 根据(W,H)和(w,h)大小比较，可以将图像的比例变换分为：拉伸和压缩 比例变换的步骤 设：输出图像中点(x,y)与输入图像中点(a,b)相对应，则它们满足以下公式：x=a*W/w,y=b*H/h f(x,y)=f(a,b) 5.1 Grey-level Interpolation “See geometric_operation.m” Problems: 拉伸时，按公式：x=a*W/w, y=b*H/h，则输出图像中有一些像素无法与输入图像中的像素对应起来。 压缩时，按公式：x=a*W/w, y=b*H/h，输入图像中多个像素点会与输出图像中的1个像素点对应。 5.1 Grey-level Interpolation Zero-Order Interpolation First-Order Interpolation Higher-Order Interpolation 5.1 Grey-level Interpolation Zero-Order (Nearest-Neighbor, 最近邻插值) Select value of input pixel “nearest” to exact location See Chap2.2.6, for example Manhattan Distance (city-block) L2 Norm (Euclidean distance, 2范数) 5.1 Grey-level Interpolation Zero-Order (Nearest-Neighbor, 最近邻插值) The steps of zero-order interpolation： Step1: 找出待求解点的邻域像素； Step2: 指定某一距离测量方法，计算待求解点与各邻域像素之间的距离； Step3: 将与待求解点距离最小的相邻像素的灰度值赋给待求解点。 5.1 Grey-level Interpolation Zero-Order (Nearest-Neighbor, 最近邻插值) Drawback Ragged (粗糙的) edges in images with straight edges 该方法最简单，插值效果尚佳，但校正后的图像有明显锯齿状，即存在灰度不连续性。 5.1 Grey-level Interpolation First-Order (Bilinear Interpolation) What is “Bilinear (双线性)”? Linear Expression “...which is of first degree in its variable (or variables).” ? y=x+1, y=x*x+x+1 This is a linear expression. 5.1 Grey-level Interpolation First-Order (Bilinear Interpolation) Bilinear (双线性) “A mathematical expression is bilinear if it is linear WRT (With Respect To, 关于) each of 2 variables or positions.” Note: “Bilinear” does not mean that it is the simple superposition (重叠) of linear expressions. 5.1 Grey-level Interpolation Consi