opencv实战之透视变换

import cv2
import numpy as np
import pytesseract
def cv_show(imgname,img):
    cv2.imshow(imgname,img)
    cv2.waitKey(0)
    cv2.destroyAllWindows()

#按左上，右上，右下，左下排序坐标
def order_points(pts):
    rec = np.zeros((4,2),dtype='float32')
    #找出左上，右下
    s = np.sum(pts,axis=1)
    rec[0] = pts[np.argmin(s)]
    rec[2] = pts[np.argmax(s)]
    #找出右上，左下 （x,y） y-x相差最小最大
    diff = np.diff(pts)
    rec[1] = pts[np.argmin(diff)]
    rec[3] = pts[np.argmax(diff)]
    return rec

#透视变换后的对应坐标
def four_point_wrap(img,pts):
    rec = order_points(pts)
    tl,tr,br,bl = rec
#找到最大的宽和高
    widthA = np.sqrt(((tr[0]-tl[0])**2)+((tr[1]-tl[1])**2))
    widthB = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
    width = max(int(widthB),int(widthA))
    lengthA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
    lengthB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
    length = max(int(lengthA),int(lengthB))
    print('width:{},length:{}'.format(width,length))
#width,length都减去100，完整的图像才会显示出来，小了的话会显示不完整，这一部分笔者也有点困惑
    dst = np.array([
        (0,0),
        (width-100,0),
        (width-100,length-100),
        (0,length-100)
    ],dtype='float32')
    M = cv2.getPerspectiveTransform(rec,dst)
    wrap = cv2.warpPerspective(img,M,(width,length))
    return wrap

def resize(img,height=None,width=None,inter=cv2.INTER_AREA):   #cv2.INIER_AREA 插值方法
    dim = None
    (h,w) = img.shape[:2]
    if width is None and height is None:
        return img
    if width is None:
        r = height / float(h)
        dim = (int(w*r),height)

    else:
        r = width / float(h)
        dim = (width,int(h*r))
    resized = cv2.resize(img,dim,interpolation=inter)
    return resized

img =  cv2.imread('ocrbook.jpg')
print(img.shape)
ratio = img.shape[0]/500.0
orig = img.copy()
#等比例缩小
img = resize(orig,height=400)
#转化成灰度图
gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
# cv_show('gray',gray)
kernel = np.array([[0, -1, 0],[-1, 5, -1],[0, -1, 0]], np.float32)
kernel1 = np.ones([3,3])
#锐度处理 对于此图像，词操作会使噪音点变多，边缘检测效果变差，故舍弃
# dst = cv2.filter2D(gray, -1, kernel=kernel)
# cv_show('dst',dst)
#高斯模糊
gaussian = cv2.GaussianBlur(gray,(1,1),0)
# cv_show('gaussian',gaussian)
#二值化处理
# thresh = cv2.threshold(gray,20,250,cv2.THRESH_BINARY|cv2.THRESH_OTSU)[1]
# cv_show('thresh',thresh)
kernel = np.ones([3,3])

#边缘检测
canny = cv2.Canny(gaussian,120,200)
cv_show('canny',canny)
#膨胀操作 使细小的未连接的边缘连接起来，使得图像有封闭区域
dilate = cv2.morphologyEx(canny,cv2.MORPH_DILATE,kernel1,iterations=2)
cv_show('dilate',dilate)
#腐蚀
erode  = cv2.morphologyEx(dilate,cv2.MORPH_ERODE,kernel1)
cv_show('erode',erode)
#轮廓检测
cnts = cv2.findContours(erode,cv2.RETR_LIST,cv2.CHAIN_APPROX_SIMPLE)[0]
#按轮廓面积升序排列
cnts = sorted(cnts,key=cv2.contourArea,reverse=True )[:4]
for cnt in cnts:
    perix = cv2.arcLength(cnt,True)
    print(perix)
    approx = cv2.approxPolyDP(cnt,0.03*perix,True)
    print('len(approx): ',len(approx))
#找到矩形
    if len(approx)==4:
        screen = approx
        break
cv2.drawContours(img,[screen],-1,(0,25,155),2)
cv_show('cnts',img)
screen = screen.reshape(4,2)*ratio
print(screen)
#透视变换
wraped = four_point_wrap(orig,screen)
cv_show('wrapwd',wraped)
#ocr检测  准确度需要训练
text = pytesseract.image_to_string(wraped)
print(text)

                                                                        

#图片来源百度图片，只为学习交流专用，如有侵权，联系笔者，立刻删除