基于opencv的PPT照片提取矫正
调研时发现了一些类似的文档矫正的工作，手动选点的工作居多。
这些工作用在真实PPT上的效果都不太好，原因可能有：（1）PPT的背景大都是浅色的，如果是屏幕显示的情况，边界比较明显，效果比较好，但对于类似新主楼教室的投影幕布而言，边界比较模糊，很难通过处理得到边界；（2）拍摄PPT的时候容易出现边界遮挡的情况；（3）PPT往往会和其他具有边界的矩形出现在同一个场景，如果只是简单的提取直线得到最大四边形，容易把PPT外部的黑板、平行的灯线一块提取出来。
实现代码如下：
import os
import cv2
import numpy as np
def preprocess(img):
# 转换为灰度图
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# 二值化
img = cv2.threshold(img, 0, 255, cv2.THRESH_BINARY | cv2.THRESH_OTSU)[1]
# 膨胀
kernel = np.ones((3, 3), np.uint8)
img = cv2.dilate(img, kernel, iterations=1)
# 腐蚀
img = cv2.erode(img, kernel, iterations=1)
return img
def preprocess_adaptive(img):
# 高斯模糊
# blurred = cv2.GaussianBlur(img, (1, 1), 0)
# 灰度化
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# 自适应阈值化
thresholded = cv2.adaptiveThreshold(gray, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 11, 2)
# 腐蚀和膨胀操作去除噪点
kernel = np.ones((3, 3), np.uint8)
eroded = cv2.erode(thresholded, kernel, iterations=1)
dilated = cv2.dilate(eroded, kernel, iterations=1)
return dilated
def find_corners(img):
"""
Finds the 4 extreme corners of the largest contour in the image.
"""
# OpenCV版本问题，cv2.findContours 版本3返回值有3个，版本4只有2个
opencv_version = cv2.__version__.split('.')[0]
if opencv_version == '3':
_, contours, _ = cv2.findContours(img.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
else:
contours, _ = cv2.findContours(img.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
# 找到最大的轮廓
contours = sorted(contours, key=cv2.contourArea, reverse=True)
polygon = contours[0]
# bottom-right point has the largest (x + y) value
# top-left has point smallest (x + y) value
# bottom-left point has smallest (x - y) value
# top-right point has largest (x - y) value
# 初始化
val_br = polygon[0][0][0] + polygon[0][0][1]
val_tl = polygon[0][0][0] + polygon[0][0][1]
val_bl = polygon[0][0][0] - polygon[0][0][1]
val_tr = polygon[0][0][0] - polygon[0][0][1]
bottom_right = polygon[0][0]
top_left = polygon[0][0]
bottom_left = polygon[0][0]
top_right = polygon[0][0]
# 寻找最大轮廓的4个边界点坐标
for i in range(1, len(polygon)):
val_add = polygon[i][0][0] + polygon[i][0][1]
val_minus = polygon[i][0][0] - polygon[i][0][1]
if val_add > val_br:
val_br = val_add
bottom_right = polygon[i][0]
if val_add < val_tl:
val_tl = val_add
top_left = polygon[i][0]
if val_minus < val_bl:
val_bl = val_minus
bottom_left = polygon[i][0]
if val_minus > val_tr:
val_tr = val_minus
top_right = polygon[i][0]
# 返回4个边界点坐标
return [top_left, top_right, bottom_right, bottom_left]
# 得到PPT图片的4个边界点坐标
def crop_and_warp(img, crop_rect):
"""
Crops and warps a rectangular section from an image into a square of similar size.
"""
# 4个边界点坐标
top_left, top_right, bottom_right, bottom_left = \
crop_rect[0], crop_rect[1], crop_rect[2], crop_rect[3]
# Explicitly set the data type to float32 or 'getPerspectiveTransform' will throw an error
# 原始坐标
src = np.array([top_left, top_right, bottom_right, bottom_left], dtype='float32')
width, height = 800, 600 # 设置矫正后的图像尺寸
dst = np.array([[0, 0], [width - 1, 0], [width - 1, height - 1], [0, height - 1],], dtype=np.float32)
# Gets the transformation matrix for skewing the image to fit a square by comparing the 4 before and after points
matrix = cv2.getPerspectiveTransform(src, dst)
# Performs the transformation on the original image
img = cv2.warpPerspective(img, matrix, (width, height))
return img
def calculate_distance(point1, point2, debug=False, threshold=50):
dis = np.sqrt((point1[0]-point2[0])**2 + (point1[1]-point2[1])**2)
if debug:
return point1 if dis > threshold else [(point1[0]+point2[0])/2, (point1[1]+point2[1])/2]
return dis
def merge_crop_rect(crop_rect, crop_rect_adaptive, threshold=50):
"""
合并两种方法得到的边界点坐标，排除异常值
"""
crop_rect_final = []
for i in range(4):
if calculate_distance(crop_rect[i], crop_rect_adaptive[i], threshold=threshold) < threshold:
crop_rect_final.append((crop_rect[i] + crop_rect_adaptive[i]) / 2)
if len(crop_rect_final) == 4:
return np.array(crop_rect_final, dtype='float32')
elif len(crop_rect_final) == 3:
crop_rect_final = []
x1, y1, x2, y2 = crop_rect[0][0], crop_rect[0][1], crop_rect[1][0], crop_rect[1][1]
x3, y3, x4, y4 = crop_rect[2][0], crop_rect[2][1], crop_rect[3][0], crop_rect[3][1]
x1_adp, y1_adp, x2_adp, y2_adp = crop_rect_adaptive[0][0], crop_rect_adaptive[0][1], crop_rect_adaptive[1][0], crop_rect_adaptive[1][1]
x3_adp, y3_adp, x4_adp, y4_adp = crop_rect_adaptive[2][0], crop_rect_adaptive[2][1], crop_rect_adaptive[3][0], crop_rect_adaptive[3][1]
dis_ab1 = calculate_distance((x1, y1), (x2, y2), threshold=threshold)
dis_ab2 = calculate_distance((x1_adp, y1_adp), (x2_adp, y2_adp), threshold=threshold)
dis_cd1 = calculate_distance((x3, y3), (x4, y4), threshold=threshold)
dis_cd2 = calculate_distance((x3_adp, y3_adp), (x4_adp, y4_adp), threshold=threshold)
if abs(dis_ab1 - dis_ab2) < threshold:
crop_rect_final.append([(x1+x1_adp)/2, (y1+y1_adp)/2])
crop_rect_final.append([(x2+x2_adp)/2, (y2+y2_adp)/2])
basic_dis = (dis_ab1 + dis_ab2) / 2
if abs(dis_cd1 - basic_dis) < threshold:
crop_rect_final.append(calculate_distance((x3, y3), (x3_adp, y3_adp), debug=True, threshold=threshold))
crop_rect_final.append(calculate_distance((x4, y4), (x4_adp, y4_adp), debug=True, threshold=threshold))
elif abs(dis_cd2 - basic_dis) < threshold:
crop_rect_final.append(calculate_distance((x3_adp, y3_adp), (x3, y3), debug=True, threshold=threshold))
crop_rect_final.append(calculate_distance((x4_adp, y4_adp), (x4, y4), debug=True, threshold=threshold))
elif abs(dis_cd1 - dis_cd2) < threshold:
basic_dis = (dis_cd1 + dis_cd2) / 2
if abs(dis_ab1 - basic_dis) < threshold:
crop_rect_final.append(calculate_distance((x1, y1), (x1_adp, y1_adp), debug=True, threshold=threshold))
crop_rect_final.append(calculate_distance((x2, y2), (x2_adp, y2_adp), debug=True, threshold=threshold))
elif abs(dis_ab2 - basic_dis) < threshold:
crop_rect_final.append(calculate_distance((x1_adp, y1_adp), (x1, y1), debug=True, threshold=threshold))
crop_rect_final.append(calculate_distance((x2_adp, y2_adp), (x2, y2), debug=True, threshold=threshold))
crop_rect_final.append([(x3+x3_adp)/2, (y3+y3_adp)/2])
crop_rect_final.append([(x4+x4_adp)/2, (y4+y4_adp)/2])
if len(crop_rect_final) == 4:
return np.array(crop_rect_final, dtype='float32')
else:
for i in range(4):
if 0 in crop_rect_adaptive[i]:
return crop_rect
return crop_rect_adaptive
if __name__ == '__main__':
file_src = 'static/src/'
file_dst = 'static/dst/'
files = os.listdir(file_src)
for file in files:
# 读取图片
img_origin = cv2.imread(file_src + file)
# img_origin = cv2.imread('static/src/test(5).jpg')
# 预处理
img = preprocess(img_origin)
adaptive = preprocess_adaptive(img_origin)
# 得到PPT图片的4个边界点坐标
crop_rect = find_corners(img)
crop_rect_adaptive = find_corners(adaptive)
# 合并两种方法得到的边界点坐标，排除异常值
height, width = img.shape[:2]
crop_rect_final = merge_crop_rect(crop_rect, crop_rect_adaptive, threshold=width / 20)
# 矫正
img = crop_and_warp(img_origin, crop_rect_final)
cv2.imshow("image", img)
cv2.waitKey(0)
cv2.destroyAllWindows()
cv2.imwrite(file_dst + file, img)
常见的出错结果如下：
最终参考上述工作的流程，选取了两种效果比较好的方法进行合并，优化后的工作效果如下：
