def adjust_aspect(img, a=1.5, show=False, binarize=False):
"""Resizes img such that the object defined by ON pixels within
has aspect ratio (height / width) a."""
if show:
showimg(img)
min_rect = cv.BoundingRect(img)
wr, hr = min_rect[2:]
w, h = cv.GetSize(img)
orig_aspect = float(hr) / wr
if orig_aspect < a:
new_w = w
new_h = int(round(a * h / orig_aspect))
else:
new_w = int(round(orig_aspect * w / a))
new_h = h
# print w, h, new_w, new_h
result = cv.CreateMat(new_h, new_w, cv.CV_8U)
cv.Resize(img, result)
if binarize:
b = cv.CreateImage((result.width, result.height), 8, 1)
# Threshold of 128 should work if the original image was
# binary.
cv.Threshold(result, b, 128, 255, cv.CV_THRESH_BINARY | cv.CV_THRESH_OTSU)
result = b
if show:
print "Original: %.2f" % orig_aspect
showimg(result)
return result
def lines(self, show=False):
result = []
line = []
line_no = self.words[0].pos[0]
if show:
cpimg = cv.CloneMat(self.img)
for w in self.words:
if w.pos[0] != line_no:
line_no = w.pos[0]
result.append(line)
line = []
if show:
print '\n'
line.append(w)
if show:
print w
xmin = min([c.bounding_box[0] for c in w.chars])
xmax = max([c.bounding_box[2] for c in w.chars])
ymin = min([c.bounding_box[1] for c in w.chars])
ymax = max([c.bounding_box[3] for c in w.chars])
cv.Rectangle(cpimg, (xmin, ymin), (xmax, ymax), color=128, thickness=2)
showimg(cpimg)
result.append(line)
return result
def make_mask(image, width, height, show=False):
x, y, w, h = cv.BoundingRect(image)
image = image[y:y + h, x:x + w]
result = cv.CreateMat(height, width, cv.CV_8U)
cv.Resize(image, result)
b = cv.CreateMat(height, width, cv.CV_8U)
cv.Threshold(result, b, 128, 1, cv.CV_THRESH_BINARY | cv.CV_THRESH_OTSU)
result = b
if show:
showimg(result)
return numpy.asarray(result)
com.xy_sym /= (h / 2) * w
return com
def __init__(self, width, height, intensity_grid_rows=3, intensity_grid_cols=3):
# Cache a copy of the mask with a border around it so we can
# find contours.
self.border_mask = cv.CreateMat(height + 2, width + 2, cv.CV_8U)
self.mask = self.border_mask[1:-1, 1:-1]
cv.Set(self.border_mask, 0)
self.offset = (-1, -1)
self.intensity = -1
self.x_sym = -1
self.y_sym = -1
self.xy_sym = -1
self.intensity_grid = np.zeros((intensity_grid_rows, intensity_grid_cols))
def xnor(x, y):
return bool(x) == bool(y)
if __name__ == '__main__':
import sys
from common.show import show as showimg
i = cv.LoadImageM(sys.argv[1], cv.CV_LOAD_IMAGE_GRAYSCALE)
coms = connected_components(i)
print len(coms)
for c in sorted(coms, key=lambda c: c.mask.rows * c.mask.cols, reverse=True)[:15]:
print c.intensity, c.x_sym, c.y_sym, c.xy_sym
showimg(c.mask)