数学之路-python计算实战(9)-机器视觉-图像插值仿射

• 插值
• Python: cv2.resize(src, dsize[, dst[, fx[, fy[, interpolation]]]]) → dst
  
• interpolation –

  interpolation method:

• INTER_NEAREST - a nearest-neighbor interpolation
• INTER_LINEAR - a bilinear interpolation (used by default)
• INTER_AREA - resampling using pixel area relation. It may be a preferred method for image decimation, as it gives moire’-free results. But when the image is zoomed, it is similar to theINTER_NEAREST method.
• INTER_CUBIC - a bicubic interpolation over 4x4 pixel neighborhood
• INTER_LANCZOS4 - a Lanczos interpolation over 8x8 pixel neighborhood

# -*- coding: utf-8 -*- 
import cv2

fn="test2.jpg"
img=cv2.imread(fn)
w=img.shape[1]  
h=img.shape[0]  

#放大,双立方插值
newimg1=cv2.resize(img,(w*2,h*2),interpolation=cv2.INTER_CUBIC)
#放大, 最近邻插值
newimg2=cv2.resize(img,(w*2,h*2),interpolation=cv2.INTER_NEAREST)
#放大, 象素关系重采样
newimg3=cv2.resize(img,(w*2,h*2),interpolation=cv2.INTER_AREA)
#缩小, 象素关系重采样
newimg4=cv2.resize(img,(300,200),interpolation=cv2.INTER_AREA)


cv2.imshow('preview1',newimg1)
cv2.imshow('preview2',newimg2)
cv2.imshow('preview3',newimg3)
cv2.imshow('preview4',newimg4)
cv2.waitKey()
cv2.destroyAllWindows()

仿射可进行缩放、旋转、平衡操作

Python:   cv2. warpAffine (src, M, dsize [, dst [, flags [, borderMode [, borderValue ] ] ] ] ) → dst

C:  void  cv WarpAffine (const CvArr*  src, CvArr*  dst, const CvMat*  map_matrix, int flags=CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS, CvScalar  fillval=cvScalarAll(0)  )

Python:   cv. WarpAffine (src, dst, mapMatrix, flags=CV_INTER_LINEAR+CV_WARP_FILL_OUTLIERS, fillval=(0, 0, 0, 0) ) → None

C:  void  cvGetQuadrangleSubPix (const CvArr*  src, CvArr*  dst, const CvMat* map_matrix )

Python:   cv. GetQuadrangleSubPix (src, dst, mapMatrix ) → None

Parameters:

src – input image. dst – output image that has the size dsize and the same type assrc . M –  transformation matrix. dsize – size of the output image. flags – combination of interpolation methods (see resize() ) and the optional flag WARP_INVERSE_MAP that means that M is the inverse transformation (  ). borderMode – pixel extrapolation method (seeborderInterpolate()); when borderMode=BORDER_TRANSPARENT , it means that the pixels in the destination image corresponding to the “outliers” in the source image are not modified by the function. borderValue – value used in case of a constant border; by default, it is 0.

The function warpAffine transforms the source image using the specified matrix:

getRotationMatrix2D

Calculates an affine matrix of 2D rotation.

C++:  Mat  getRotationMatrix2D (Point2f  center, double  angle, double  scale )

Python:   cv2. getRotationMatrix2D (center, angle, scale ) → retval

C:  CvMat*  cv2DRotationMatrix (CvPoint2D32f  center, double  angle, double  scale, CvMat*  map_matrix )

Python:   cv. GetRotationMatrix2D (center, angle, scale, mapMatrix ) → None

Parameters:	center – Center of the rotation in the source image. angle – Rotation angle in degrees. Positive values mean counter-clockwise rotation (the coordinate origin is assumed to be the top-left corner). scale – Isotropic scale factor. map_matrix – The output affine transformation, 2x3 floating-point matrix.

The function calculates the following matrix:

where

The transformation maps the rotation center to itself. If this is not the target, adjust the shift.

仿射变换，又称仿射映射，是指在几何中，一个向量空间进行一次线性变换并接上一个平移，变换为另一个向量空间。

一个对向量 平移，与旋转放大缩小 的仿射映射为

上式在 齐次坐标上，等价于下面的式子

为了表示仿射变换，需要使用齐次坐标，即用三维向量 (x, y, 1) 表示二维向量，对于高维来说也是如此。按照这种方法，就可以用矩阵乘法表示变换。 ;  变为

在矩阵中增加一列与一行，除右下角的元素为 1 外其它部分填充为 0，通过这种方法，所有的线性变换都可以转换为仿射变换。例如，上面的旋转矩阵变为

通过这种方法，使用与前面一样的矩阵乘积可以将各种变换无缝地集成到一起

# -*- coding: utf-8 -*- 
import cv2

fn="test3.jpg"
img=cv2.imread(fn)
w=img.shape[1]  
h=img.shape[0]  
#得到仿射变换矩阵，完成旋转
#中心
mycenter=(h/2,w/2)
#旋转角度
myangle=90
#缩放尺度
myscale=0.5
#仿射变换完成缩小并旋转
transform_matrix=cv2.getRotationMatrix2D(mycenter,myangle,myscale)

newimg=cv2.warpAffine(img,transform_matrix,(h,w))
cv2.imshow('preview',newimg)

cv2.waitKey()
cv2.destroyAllWindows()

本博客所有内容是原创，如果转载请注明来源

http://blog.csdn.net/myhaspl/

本博客所有内容是原创，如果转载请注明来源

http://blog.csdn.net/myhaspl/