dsim

import numpy as np
import matplotlib.pyplot as plt
import cv2
import os
import urllib2
import sys

def load_sample_image(filename):
   response = urllib2.urlopen('https://raw.github.com/Itseez/opencv/master/samples/c/' + filename)
   return cv2.imdecode(np.fromstring(response.read(), dtype='uint8'), 0)

def show_image(img):
    plt.imshow(img)
    if len(img.shape)<3:
        plt.gray()
    plt.axis("off")
    #plt.show()
    plt.savefig('sss.png', dpi=900)

img11 = cv2.imread(sys.argv[2]) #load_sample_image('box.png')
img1 = cv2.cvtColor( img11, cv2.COLOR_RGB2GRAY)
img22 = cv2.imread(sys.argv[3]) #load_sample_image('box_in_scene.png')
img2 = cv2.cvtColor( img22, cv2.COLOR_RGB2GRAY)

print("Read img1 of size {} and img2 of size {}.".format(img1.shape, img2.shape))
print("Image 1:")
#show_image(img1)
print("Image 2:")
#show_image(img2)

th = int(sys.argv[1])

octaves = 7
oct_layers = 5

detector = cv2.SURF(th,octaves,oct_layers)
norm = cv2.NORM_L2
matcher = cv2.BFMatcher(norm)

kp1, desc1 = detector.detectAndCompute(img1, None)
kp2, desc2 = detector.detectAndCompute(img2, None)

print '%s feature points for image1' % len(kp1)
print '%s feature points for image2' % len(kp2)

raw_matches = matcher.knnMatch(desc1, trainDescriptors = desc2, k = 2)

def filter_matches(kp1, kp2, matches, ratio = 0.9):
    """
    Keep only matches that have distance ratio to
    second closest point less than 'ratio'.
    """
    mkp1, mkp2 = [], []
    for m in matches:
        if m[0].distance < m[1].distance * ratio:
            m = m[0]
            mkp1.append( kp1[m.queryIdx] )
            mkp2.append( kp2[m.trainIdx] )
    p1 = np.float32([kp.pt for kp in mkp1])
    p2 = np.float32([kp.pt for kp in mkp2])
    kp_pairs = zip(mkp1, mkp2)
    return p1, p2, kp_pairs

p1, p2, kp_pairs = filter_matches(kp1, kp2, raw_matches)

H, status = cv2.findHomography(p1, p2, cv2.RANSAC, 40.0)

h1, w1 = img1.shape[:2]
h2, w2 = img2.shape[:2]
vis = np.zeros((max(h1, h2), w1), np.uint8)
vis[:h1, :w1] = img1
#vis[:h2, w1:w1+w2] = img2
vis = cv2.cvtColor(vis, cv2.COLOR_GRAY2BGR)

#show_image(vis)

p1 = np.int32([kpp[0].pt for kpp in kp_pairs])
p2 = np.int32([kpp[1].pt for kpp in kp_pairs]) #+ (w1, 0)

# plot the matches
color = (0, 255, 0)
ch_valid = 0
for (x1, y1), (x2, y2), inlier in zip(p1, p2, status):
    if inlier:
        if (abs(x2-x1)>3 or abs(y2-y1)>3):
            cv2.circle(vis, (x1, y1), 2, color, -1)
            #cv2.circle(vis, (x2, y2), 2, color, -1)
            cv2.line(vis, (x1, y1), (x2, y2), color)
            ch_valid = ch_valid + 1
print '%s number of valid vectors' % ch_valid
show_image(vis)