这是一个经典的情况dilate https://opencv24-python-tutorials.readthedocs.io/en/latest/py_tutorials/py_imgproc/py_morphological_ops/py_morphological_ops.html#dilation。这个想法是相邻的文本对应于同一问题,而较远的文本是另一个问题的一部分。每当您想要将多个项目连接在一起时,您可以扩大它们以将相邻轮廓连接成单个轮廓。这是一个简单的方法:
-
获取二值图像。 加载图像 https://www.geeksforgeeks.org/python-opencv-cv2-imread-method/, 转换成灰度 https://opencv24-python-tutorials.readthedocs.io/en/stable/py_tutorials/py_imgproc/py_colorspaces/py_colorspaces.html, 高斯模糊 https://opencv24-python-tutorials.readthedocs.io/en/latest/py_tutorials/py_imgproc/py_filtering/py_filtering.html#gaussian-filtering, then 大津的门槛 https://opencv24-python-tutorials.readthedocs.io/en/latest/py_tutorials/py_imgproc/py_thresholding/py_thresholding.html#otsus-binarization以获得二值图像。
-
消除小噪音和伪影。我们创建一个矩形核 https://opencv24-python-tutorials.readthedocs.io/en/latest/py_tutorials/py_imgproc/py_morphological_ops/py_morphological_ops.html#structuring-element and 变形开放 https://opencv24-python-tutorials.readthedocs.io/en/latest/py_tutorials/py_imgproc/py_morphological_ops/py_morphological_ops.html#opening去除图像中的小噪声和伪影。
-
将相邻的单词连接在一起。我们创建一个更大的矩形内核并且dilate https://opencv24-python-tutorials.readthedocs.io/en/latest/py_tutorials/py_imgproc/py_morphological_ops/py_morphological_ops.html#dilation将各个轮廓合并在一起。
-
检测问题。从这里我们找到轮廓 https://docs.opencv.org/2.4/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html#findcontours,使用从上到下对轮廓进行排序imutils.sort_contours() https://github.com/PyImageSearch/imutils/blob/master/imutils/contours.py#L7,用过滤器最小轮廓面积 https://docs.opencv.org/2.4/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html#contourarea,得到矩形边界矩形坐标 https://docs.opencv.org/2.4/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=boundingrect#boundingrect and 突出显示矩形轮廓 https://opencv24-python-tutorials.readthedocs.io/en/latest/py_tutorials/py_imgproc/py_contours/py_contours_begin/py_contours_begin.html。然后,我们使用 Numpy 切片裁剪每个问题并保存 ROI 图像。
获得二值图像的大津阈值
这是有趣的部分发生的地方。我们假设相邻的文本/字符是同一问题的一部分,因此我们将各个单词合并成一个轮廓。问题是靠近在一起的单词的一部分,因此我们扩展以将它们连接在一起。
个别问题以绿色突出显示
热门问题
底部问题
已保存的 ROI 问题(假设从上到下)
Code
import cv2
from imutils import contours
# Load image, grayscale, Gaussian blur, Otsu's threshold
image = cv2.imread('1.png')
original = image.copy()
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
blur = cv2.GaussianBlur(gray, (7,7), 0)
thresh = cv2.threshold(blur, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)[1]
# Remove small artifacts and noise with morph open
open_kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (5,5))
opening = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, open_kernel, iterations=1)
# Create rectangular structuring element and dilate
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (9,9))
dilate = cv2.dilate(opening, kernel, iterations=4)
# Find contours, sort from top to bottom, and extract each question
cnts = cv2.findContours(dilate, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0] if len(cnts) == 2 else cnts[1]
(cnts, _) = contours.sort_contours(cnts, method="top-to-bottom")
# Get bounding box of each question, crop ROI, and save
question_number = 0
for c in cnts:
# Filter by area to ensure its not noise
area = cv2.contourArea(c)
if area > 150:
x,y,w,h = cv2.boundingRect(c)
cv2.rectangle(image, (x, y), (x + w, y + h), (36,255,12), 2)
question = original[y:y+h, x:x+w]
cv2.imwrite('question_{}.png'.format(question_number), question)
question_number += 1
cv2.imshow('thresh', thresh)
cv2.imshow('dilate', dilate)
cv2.imshow('image', image)
cv2.waitKey()