def main(num=18, name='prtn', focal_name='pano.txt', output_name='out.jpg'):
imgs_cylindral = CC(num, name, focal_name)
#convert to cylindral coordinate
imgs_cylindral_gray = ToGray(imgs_cylindral) #to gray scale
total_shift_list = []
panorama_h_len = 0
panorama_w_len = 0
for i in xrange(len(imgs_cylindral_gray) - 1): #pairwise matching
feature_points1 = feature.harris_detector(imgs_cylindral_gray[i])
feature_description1 = feature.harris_descriptor(
imgs_cylindral_gray[i], feature_points1)
feature_points2 = feature.harris_detector(imgs_cylindral_gray[i + 1])
feature_description2 = feature.harris_descriptor(
imgs_cylindral_gray[i + 1], feature_points2)
matches = feature.matching(feature_description1, feature_description2)
match_points1 = feature_points1[matches[:, 0],
0:2] #only xy coordinate
match_points2 = feature_points2[matches[:, 1], 0:2]
match_list = np.hstack((match_points1, match_points2))
[shift_x, shift_y, matches_ransac] = RANSAC(match_list)
total_shift_list.append([shift_x, shift_y])
total_shift_list = np.array(total_shift_list).astype(int)
print 'total_shift_list size:', len(total_shift_list) #should be num - 1
print total_shift_list #shift_x is horizontal direction
panorama_h_len = imgs_cylindral[-1].shape[0]
panorama_w_len = imgs_cylindral[-1].shape[1]
'''
start stitch
'''
panorama_w_len += int(np.ceil(np.sum(total_shift_list, axis=0)[0]))
panorama_h_len += abs(int(2 *
np.ceil(np.sum(total_shift_list, axis=0)[1])))
print 'start stitching, panorama size(h,w):', panorama_h_len, panorama_w_len
panorama = np.zeros(
(panorama_h_len, panorama_w_len, 3)) #determine size of panorama
accumulate_shift_w = 0
accumulate_shift_h = -abs(int(np.ceil(np.sum(total_shift_list,
axis=0)[1])))
overlap_width = 0
#test = np.zeros((800,800,3))
for i in xrange(len(imgs_cylindral)):
if i >= 1:
overlap_width += imgs_cylindral[i - 1].shape[1] - total_shift_list[
i - 1][0]
for j in xrange(int(imgs_cylindral[i].shape[0])):
for k in xrange(int(imgs_cylindral[i].shape[1])):
if np.array_equal(
panorama[accumulate_shift_h - j,
accumulate_shift_w - k], np.zeros(3)):
panorama[accumulate_shift_h - j, \
accumulate_shift_w - k, :] = imgs_cylindral[i][-j, -k, :]
'''
if i == 1:
test[-j, -k, :] = imgs_cylindral[i][-j, -k, :]
'''
elif np.array_equal(imgs_cylindral[i][-j, -k], np.zeros(3)):
continue
else:
if overlap_width == 0:
print i, k, overlap_width
assert overlap_width != 0
t = k / float(overlap_width)
if not t <= 1:
print i, k, overlap_width
assert t <= 1
panorama[accumulate_shift_h - j, \
accumulate_shift_w - k, :] = t*imgs_cylindral[i][-j, -k, :] + (1-t)*panorama[accumulate_shift_h - j, accumulate_shift_w - k, :]
#test[-j, -k, :] = t*imgs_cylindral[i][-j, -k, :]
#print k, overlap_width, t
if i == len(imgs_cylindral) - 1:
break
accumulate_shift_h -= total_shift_list[i][
1] #I think that vertical shift should be less than 4 pixels
accumulate_shift_w -= total_shift_list[i][0]
imsave(output_name, panorama)
#imsave('t.jpg', test)
print 'finish'
corner_response2,
kernel=const.DESCRIPTOR_SIZE,
threshold=const.FEATURE_THRESHOLD)
print(str(len(descriptors2)) + ' features extracted.')
cache_feature = [descriptors2, position2]
if const.DEBUG:
cv2.imshow('cr1', corner_response1)
cv2.imshow('cr2', corner_response2)
cv2.waitKey(0)
print(' - Feature matching .... ', end='', flush=True)
matched_pairs = feature.matching(descriptors1,
descriptors2,
position1,
position2,
pool,
y_range=const.MATCHING_Y_RANGE)
print(str(len(matched_pairs)) + ' features matched.')
if const.DEBUG:
utils.matched_pairs_plot(img1, img2, matched_pairs)
print(' - Finding best shift using RANSAC .... ', end='', flush=True)
shift = stitch.RANSAC(matched_pairs, shifts[-1])
shifts += [shift]
print('best shift ', shift)
print(' - Stitching image .... ', end='', flush=True)
stitched_image = stitch.stitching(stitched_image,
img2,