Python3 skimage - count number of pictures

Question

I have an image which I'm using skimage to try and detect:

1. how many images are actually within the page - I'm expecting it to 'count' 5

2. find the corners of each image - so if it counts 5 above our maxCorners should be 5*4=20

3. draw straight lines between each corner and 'mask' each of the 5 images

right now all I got is the image being read, doing a fill holes and thats about it - guidance on the rest?

from scipy import ndimage as nd
import imageio
from skimage import io, filters
from scipy import ndimage
import matplotlib.pyplot as plt

filename = "C:\\Users\\Tony\\Pictures\\img807.tif"

im = io.imread(filename, as_gray=True)
val = filters.threshold_otsu(im)
drops = ndimage.binary_fill_holes(im < val)
plt.imshow(drops, cmap='gray')
plt.show()

I've tried looking at these resources:

• https://scikit-image.org/docs/dev/auto_examples/color_exposure/plot_regional_maxima.html#sphx-glr-auto-examples-color-exposure-plot-regional-maxima-py

• https://scikit-image.org/docs/dev/auto_examples/edges/plot_convex_hull.html#sphx-glr-auto-examples-edges-plot-convex-hull-py

• https://scikit-image.org/docs/dev/auto_examples/features_detection/plot_corner.html

Specifically the last one corner detection...

Here's the original image:

True source link (12hrs): https://u.pcloud.link/publink/show?code=XZDONYXZVkUNFT2qcEFk4nFYYnx7d8swzaD7

Answer 1

This Answer was the key to solve this problem.

Coords:

[[(38, 11), (251, 364)], [(254, 62), (592, 266)], [(254, 312),

(592, 518)], [(46, 456), (247, 797)], [(346, 557), (526, 797)]]

import numpy as np
import matplotlib.pyplot as plt
import cv2
import itertools

#====================================================
img = cv2.imread('input.jpg', 0)
blur = cv2.blur(img,(3,3))

blur[blur>225] = 0

sobelx = cv2.Sobel(blur,cv2.CV_64F,1,0,ksize=5)
sobely = cv2.Sobel(blur,cv2.CV_64F,0,1,ksize=5)
sobel = np.sqrt( sobelx**2 + sobely**2)

sobel = (255 * sobel)/(sobel.max() - sobel.min())
sobel = sobel.astype(np.uint8)
sobel[sobel<20] = 0
sobel[sobel>20] = 255
#====================================================

_,thresh = cv2.threshold(blur,127,255,1)
thresh = thresh + sobel

median = cv2.medianBlur(thresh,3)

gray_scale = median.copy()

image = np.stack([img, img, img], axis=2)

img_bin = cv2.Canny(gray_scale,50,110)
dil_kernel = np.ones((3,3), np.uint8)
img_bin=cv2.dilate(img_bin,dil_kernel,iterations=1)

line_min_width = 7

kernal_h = np.ones((2,line_min_width), np.uint8)
img_bin_h = cv2.morphologyEx(img_bin, cv2.MORPH_OPEN, kernal_h)

kernal_v = np.ones((line_min_width,1), np.uint8)
img_bin_v = cv2.morphologyEx(img_bin, cv2.MORPH_OPEN, kernal_v)

img_bin_final=img_bin_h|img_bin_v
final_kernel = np.ones((3,3), np.uint8)
img_bin_final=cv2.dilate(img_bin_final,final_kernel,iterations=1)

_, _, stats, _ = cv2.connectedComponentsWithStats(~img_bin_final, connectivity=8, ltype=cv2.CV_32S)

coords = []
### 1 and 0 and the background and residue connected components whihc we do not require
for x,y,w,h,area in stats[2:]:
    if area>15000:
        coords.append([(x,y),(x+w,y+h)])

def bb_intersection(coords, boxA, boxB):
    # determine the (x, y)-coordinates of the intersection rectangle
    xA = max(boxA[0][0], boxB[0][0])
    yA = max(boxA[0][1], boxB[0][1])
    xB = min(boxA[1][0], boxB[1][0])
    yB = min(boxA[1][1], boxB[1][1])
    # compute the area of intersection rectangle
    interArea = max(0, xB - xA + 1) * max(0, yB - yA + 1)
    # compute the area of both rectangles
    boxAArea = (boxA[1][0] - boxA[0][0] + 1) * (boxA[1][1] - boxA[0][1] + 1)
    boxBArea = (boxB[1][0] - boxB[0][0] + 1) * (boxB[1][1] - boxB[0][1] + 1)

    if(interArea == boxAArea):
        coords.remove(boxA)
    elif(interArea == boxBArea):
        coords.remove(boxB)
#
for boxa, boxb in itertools.combinations(coords, 2):
    bb_intersection(coords, boxa, boxb)

for coord in coords:
    cv2.rectangle(image,coord[0],coord[1],(0,255,0),1)

print(coords)

plt.imshow(image)
plt.title("There are {} images".format(len(coords)))
plt.axis('off')
plt.show()

Edit:

This Answer isn't a general solution, the parameters has to be tuned accordingly, for the original image change this block, and the code will work perfectly:

img = cv2.imread('input.tif', 0)

img = cv2.resize(img, (605, 830))

blur = img.copy()

blur[blur>225] = 0

sobelx = cv2.Sobel(blur,cv2.CV_64F,1,0,ksize=3)
sobely = cv2.Sobel(blur,cv2.CV_64F,0,1,ksize=3)
sobel = np.sqrt( sobelx**2 + sobely**2)

sobel = (255 * sobel)/(sobel.max() - sobel.min())
sobel = sobel.astype(np.uint8)
sobel[sobel<40] = 0
sobel[sobel>40] = 255

Coords:

[[(31, 16), (240, 364)], [(253, 56), (600, 265)], [(254, 309),

(605, 520)], [(40, 456), (248, 803)], [(347, 557), (534, 803)]]