Python, cv2, numpy to find images in an image

Question

Similar to the question here, Python, arranging sequences in 2 combined arrays, it’s interesting to have a solution, to find out how many different images and their sequences in a combination.

The target image is:

c:\four.jpg

By using “cv2” and “numpy”, what’s the way to have a result like:

['tooth fairy', 'santa and deer', 'santa', 'deer', 'tooth fairy', 'deer', 'santa and deer', 'santa', 'tooth fairy', 'santa']

The individuals are:

c:\tooth fairy.jpg

c:\santa and deer.jpg

c:\santa.jpg

c:\deer.jpg

thanks.

Answer 1

Interesting example. Sure I'll condensate what we got in the chat.

First open your images in rgb space with imread, then you can use matchtemplate and numpy.where to find the position, and from the y coordinate obtain the sequence, as

import cv2
import numpy as np

img_rgb = cv2.imread("four.jpg") 

template = cv2.imread('fairy.jpg') 
template1 = cv2.imread('sad.jpg')
template2 = cv2.imread('san.jpg')
template3 = cv2.imread('deer.jpg')

res = cv2.matchTemplate(img_rgb,template,cv2.TM_CCOEFF_NORMED) 
res1 = cv2.matchTemplate(img_rgb,template1,cv2.TM_CCOEFF_NORMED) 
res2 = cv2.matchTemplate(img_rgb,template2,cv2.TM_CCOEFF_NORMED) 
res3 = cv2.matchTemplate(img_rgb,template3,cv2.TM_CCOEFF_NORMED) 

threshold = 0.99 

loc = np.where (res >= threshold) 
loc1 = np.where (res1 >= threshold) 
loc2 = np.where (res2 >= threshold)
loc3 = np.where (res3 >= threshold)

fairy = list(loc[0]) 
sandeer = list(loc1[0]) 
san = list(loc2[0]) 
deer = list(loc3[0]) 

x=sorted(fairy+sandeer+san+deer) 
out=', '.join(['tooth fairy' if y in fairy else 'santa and deer' if y in sandeer else 'santa' if y in san else 'deer' for y in x])

print out