无法使用 Python 分割图像

Question

我需要将我的图像拆分或裁剪成图像中显示的几个框。下面给出了我的代码，它可以拆分图像，但我无法使用我的代码创建 15 个不同的框。

我的代码如下：

import math, cv2
from scipy import misc
import numpy

def getFactors(num):
    """
    Split the input number into factors nearest to its square root. May not be
    the most efficient for large numbers, but will do for numbers smaller than 1000.
    """
    sqt = int(math.sqrt(num))
    if (num % sqt) == 0:
        return (sqt,int(num/sqt))

    num1 = sqt
    num2 = sqt
    while True:
        num1 += 1
        num2 -= 1
        if (num1 >= num) or (num2 <= 0):
            return (num, 1)
        if (num % num1) == 0:
            return (num1, int(num/num1))
        if (num % num2) == 0:
            return (num2, int(num/num2))
    return

def splitImage(img, numsplits):
    """
    Split the input image into number of splits provided by the second argument.
    The results are stored in a numpy array res and returned. The last index of the
    res array indexes the individual parts.
    """
    # Get the factors for splitting. So if the number of splits is 9, then (3,3)
    # or if 6 then (2,3) etc.
    factors = getFactors(numsplits)
    # Height and width of each split
    h = int(img.shape[0] / factors[0])
    w = int(img.shape[1] / factors[1])
    # Handle both color and B&W images
    if img.ndim >= 3:
        size = (h,w,img.shape[2],numsplits)
    else:
        size = (h,w,numsplits)
    # Initialize the result array
    res = numpy.ndarray( size, dtype = img.dtype )
    # Iterate through the number of factors to split the source image horizontally
    # and vertically, and store the resultant chunks
    for i in range(factors[0]):
        for j in range(factors[1]):
            if img.ndim >= 3:
                res[:,:,:,((i*factors[1])+j)] = img[(i*h):((i+1)*h), (j*w):((j+1)*w),:]
            else:
                res[:,:,((i*factors[1])+j)] = img[(i*h):((i+1)*h), (j*w):((j+1)*w)]

    return res

def cropImage(img):
    """
    Detect lines in the image to crop it so that the resultant image can be split well.
    We use here Canny edge detection followed by Hough Line Transform.
    """
    # Convert image to grayscale
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    # Detect edges and lines
    edges = cv2.Canny(gray, 50, 150, apertureSize = 3)
    lines = cv2.HoughLines(edges, 1, numpy.pi/90, 200)

    min_x = img.shape[0]
    max_x = 0
    min_y = img.shape[1]
    max_y = 0
    # Find the extremal horizontal and vertical coordinates to crop
    for i in range(len(lines[:,0,0])):
        rho = lines[i,0,0]
        theta = lines[i,0,1]
        a = numpy.cos(theta)
        b = numpy.sin(theta)
        x = a*rho
        y = b*rho

        if abs(a) < 1e-06 :
            if min_y > int(y):
                min_y = int(y)
            if max_y < int(y):
                max_y = int(y)
        if abs(b) < 1e-06 :
            if min_x > int(x):
                min_x = int(x)
            if max_x < int(x):
                max_x = int(x)

    return img[min_y:max_y, min_x:max_x, :]

# Read image     
img = misc.imread('tmp.png')
# Crop the image
img = cropImage(img)
# Call the splitter function
res = splitImage(img, 6)
# Save the results to files
for i in range(res.shape[-1]):
    if img.ndim >= 3:
        misc.imsave('res_{0:03d}.png'.format(i),res[:,:,:,i])
    else:
        misc.imsave('res_{0:03d}.png'.format(i),res[:,:,i])

现在我想将其裁剪或拆分为 15 个不同的框，可在此处附加的图像中找到。我找不到任何解决方案。

Answer 1

OpenCV：查找轮廓

一个简单的解决方案是使用 OpenCV findContours 的内置函数

它将找到您图像中存在的所有轮廓（以轮廓列表的形式返回），在您的示例中，它很有可能会检测到您的框。
然后您可以遍历此列表以获取轮廓的边界框。从那里您可以通过使用边界框坐标对原始图像进行切片来创建子图像。

_, contours, _ = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
for contour in countours:
    (_x, _y, w, h) = cv2.boundingRect(contour)

cv2.findContours使用官方教程：
https://docs.opencv.org/3.3.1/d4/d73/tutorial_py_contours_begin.html