This is a very simple python library to blend signature images with given background images, it is heavily inspired in the code available here and uses the blending process described in [1].

To make sure all the available functions are working properly on your system run all the test functions available with the make command.

The library is composed of three modules, two for processing the signature images and one for blending it with background images.

• color module
• transform module
• blend module
• measure module

the color module contains functions to binarize and remove the background of a signature image with a whitish background. To binarize it uses posterization just like describe in [1].

sigback.processing.color.binarize (img, nl, method='postarization')

Binarizes an image.

Parameters

sig (M, N)ndarray
Grayscale image representing the signature. The pixel values must be between 0 and 1.

nl int
Posterization level used.

method str
Thresholding method to be used, so far only the 'posterization' option is implemented.

example

from skimage.io import imread
from skimage import img_as_float64

from sigback.processing import color


signature = imread('some_signature.png', as_gray=True)
signature = img_as_float64(signature)

bin_signature = color.binarize(signature)

sigback.processing.color.remove_background(img, nl=3, method='posterization')

Removes the background of a grayscale signature image. This method makes a call to the binarize method described above.

Parameters

sig (M, N)ndarray
Grayscale image representing the signature. The pixel values must be between 0 and 1.

nl int
Posterization level used to threshold the image.

method str
Thresholding method to be used, so far only the 'posterization' option is implemented.

Use example

from skimage.io import imread
from skimage import img_as_float64

from sigback.processing import color


signature = imread('some_signature.png', as_gray=True)
signature = img_as_float64(signature)

no_bg_signature = color.remove_background(signature)

The transform module contains functions to remove the borders of a signature image and to calculate the barycenter of a signature. It is important to notice that to this functions work correctly the signature image must have a perfectly white background, this can be achieved by calling the functions of the color module or making your own background removal functions.

sigback.processing.transform.remove_border(img)

Removes the border of an image with white background.

Parameters

img (N, M)ndarray
Grayscale image with pixel values between 0 and 1.

use example

from skimage.io import imread
from skimage import img_as_float64

from sigback.processing import transform  


signature = imread('other_signature.png', as_gray=True)
signature = img_as_float64(sisgnature)

no_border_signature = transform.remove_border(signature)

sigback.processing.transform.barycenter(img)

Calculates the barycenter of an image with a white background.

Parameters

img (N, M)ndarray
Grayscale image with pixel values between 0 and 1.

use example

from skimage.io import imread
from skimage import img_as_float64

from sigback.processing import transform  


signature = imread('other_signature.png', as_gray=True)
signature = img_as_float64(sisgnature)

signature_barycenter = transform.barycenter(signature)

The blend module contains functions to blend a signature image to a background image and to blend all signatures from one directory into all backgrounds on another directory.

sigback.blend.blend.blend(sig, sig_barycenter, doc, doc_center, keep_size=False)

Blends a signature image to a background image.

Parameters

sig (N, M)ndarray
Grayscale image with pixel values between 0 and 1 and whitish background.

sig_barycenter tuple
The barycenter of the signature image, if the barycenter is not known (and the method described above cannot be used) it is possible to pass (0, 0), with this the signatures will be blent together, but the position of the signature in the background will be the top left.

doc (N, M)ndarray
Grayscale image with pixel values between 0 and 1. This will be used as the background of the blended image.

doc_center tuple
Center of the area reserved for the signature in the document. This is useful if the background image is of a document, otherwise (0, 0) can be passed to this parameter and the images will be blended ignoring the position of the signature on the background.

keep_size bool
If True the blended signature will be of the same size as the source signature.

use example

from skimage.io import imread
from skimage import img_as_float64

from sigback.blend import blend


background = imread('background.png', as_gray=True)
background = img_as_float64(background)

signature = imread('third_signature.png', as_gray=True)
signature = img_as_float64(signature)

blended = blend.blend(sig, (200, 100), background, (800, 900))
blended_no_placement = blend.blend(sig, (0, 0), background, (0, 0))

sigback.blend.blend.blend_dirs(sigs_dir, docs_dir, doc_centers, save_dir, keep_size=False, seed=None)

Blends all signatures from a directory with the backgrounds on another directory.

Parameters

sigs_dir str
The path to the directory containing all the signatures.

docs_dir str
The path to the directory containing all the documents (or any other kind of background).

doc_centers list
List of tuples containing the centers of the documents, if the center is not known (0, 0) can be passed. The list must be in the same order as the documents would be if they were sorted alphabetically

save_dir str
The path to the directory in which the blended images will be saved.

keep_size bool
If True the blended signature will be of the same size as the source signature. This will be applied to all signatures in the source directory.

seed int
Used to randomly shuffle the signatures, should be only used for testing or reproducibility.

Use example

import os

from sigback.blend.blend import blend_dirs

cx = [
    1800, # 1,png
    2164, # 10.png
    2890, # 11.png
    2264, # 12.png
    2300, # 13.png
    2055, # 14.png
    2785, # 15.png
    2586, # 16.png
    1286, # 17.png
    1285, # 18.png
    2472, # 19.png
    2602, # 2.png
    699, # 20.png
    2067, # 3.png
    2713, # 4.png
    2077, # 5.png
    2638, # 6.png
    2287, # 7.png
    2195, # 8.png
    2071 # 9.png
]
cy = [
    1215, # 1.png
    836, # 10.png
    1189, # 11.png
    1140, # 12.png
    783, # 13.png
    1807, # 14.png
    1479, # 15.png
    452, # 16.png
    393, # 17.png
    376, # 18.png
    1877, # 19.png
    659, # 2.png
    683, # 20.png
    979, # 3.png
    1269, # 4.png
    834, # 5.png
    1087, # 6.png
    965, # 7.png
    767, # 8.png
    805 # 9.png
]

sigs_dir = '/home/heitor/data/ic/sigver/data/UTSig_Crop/'
docs_dir = '/home/heitor/data/ic/image-blender/checks/all_checks'
doc_centers = list(zip(cy, cx))
save_dir = '/home/heitor/data/ic/sigver/data/background_utsig/'

blend_dirs(sigs_dir, docs_dir, doc_centers, save_dir)

The measure module contains functions to measure properties of images, like maximum and minimum size.

sigback.processing.measure.minmax(directory)

Gets the maximum and the minimum number of rows of the images in a given directory.

Parameters

directory str
The path to the directory that contains the images. The directory must contain only the images.

use example

from skimage import img_as_float64

from sigback.processing import measure


directory = '/tmp/images'
result = measure.minmax(directory) # result = (150, 300, 100, 200)

[1] Miguel A. Ferrer, Francisco Vargas, Aythami Morales, Aaron Ordo\F1ez, "Robustness of Off-line Signature Verification based on Gray Level Features", IEEE Transactions on Information Forensics and Security, vol. 7, no. 3, pp. 966-977, June 2012.