A series of convenience functions to make basic image processing functions such as translation, rotation, resizing, skeletonization, and displaying Matplotlib images easier with OpenCV and Python.
For more information, along with a detailed code review check out the following posts on the PyImageSearch.com blog:
- http://www.pyimagesearch.com/2015/02/02/just-open-sourced-personal-imutils-package-series-opencv-convenience-functions/
- http://www.pyimagesearch.com/2015/03/02/convert-url-to-image-with-python-and-opencv/
- http://www.pyimagesearch.com/2015/04/06/zero-parameter-automatic-canny-edge-detection-with-python-and-opencv/
- http://www.pyimagesearch.com/2014/09/01/build-kick-ass-mobile-document-scanner-just-5-minutes/
Provided you already have NumPy, Matplotlib, and OpenCV already installed, the imutils package is completely pip-installable:
$ pip install imutils
Translation is the shifting of an image in either the x or y direction. To translate an image in OpenCV you would need to supply the (x, y)-shift, denoted as (tx, ty) to construct the translation matrix M:
And from there, you would need to apply the cv2.warpAffine function.
Instead of manually constructing the translation matrix M and calling cv2.warpAffine, you can simply make a call to the translate function of imutils.
# translate the image x=25 pixels to the right and y=75 pixels up translated = imutils.translate(workspace, 25, -75)
<img src="docs/images/translation.png?raw=true" alt="Translation example"/ style="max-width: 500px;">
Rotating an image in OpenCV is accomplished by making a call to cv2.getRotationMatrix2D and cv2.warpAffine. Further care has to be taken to supply the (x, y)-coordinate of the point the image is to be rotated about. These calculation calls can quickly add up and make your code bulky and less readable. The rotate function in imutils helps resolve this problem.
# loop over the angles to rotate the image
for angle in xrange(0, 360, 90):
# rotate the image and display it
rotated = imutils.rotate(bridge, angle=angle)
cv2.imshow("Angle=%d" % (angle), rotated)
<img src="docs/images/rotation.png?raw=true" alt="Rotation example"/ style="max-width: 500px;">
Resizing an image in OpenCV is accomplished by calling the cv2.resize function. However, special care needs to be taken to ensure that the aspect ratio is maintained. This resize function of imutils maintains the aspect ratio and provides the keyword arguments width and height so the image can be resized to the intended width/height while (1) maintaining aspect ratio and (2) ensuring the dimensions of the image do not have to be explicitly computed by the developer.
Another optional keyword argument, inter, can be used to specify interpolation method as well.
# loop over varying widths to resize the image to
for width in (400, 300, 200, 100):
# resize the image and display it
resized = imutils.resize(workspace, width=width)
cv2.imshow("Width=%dpx" % (width), resized)
<img src="docs/images/resizing.png?raw=true" alt="Resizing example"/ style="max-width: 500px;">
Skeletonization is the process of constructing the "topological skeleton" of an object in an image, where the object is presumed to be white on a black background. OpenCV does not provide a function to explicitly construct the skeleton, but does provide the morphological and binary functions to do so.
For convenience, the skeletonize function of imutils can be used to construct the topological skeleton of the image.
The first argument, size is the size of the structuring element kernel. An optional argument, structuring, can be used to control the structuring element -- it defaults to cv2.MORPH_RECT , but can be any valid structuring element.
# skeletonize the image
gray = cv2.cvtColor(logo, cv2.COLOR_BGR2GRAY)
skeleton = imutils.skeletonize(gray, size=(3, 3))
cv2.imshow("Skeleton", skeleton)
<img src="docs/images/skeletonization.png?raw=true" alt="Skeletonization example"/ style="max-width: 500px;">
In the Python bindings of OpenCV, images are represented as NumPy arrays in BGR order. This works fine when using the cv2.imshow function. However, if you intend on using Matplotlib, the plt.imshow function assumes the image is in RGB order. A simple call to cv2.cvtColor will resolve this problem, or you can use the opencv2matplotlib convenience function.
# INCORRECT: show the image without converting color spaces
plt.figure("Incorrect")
plt.imshow(cactus)
# CORRECT: convert color spaces before using plt.imshow
plt.figure("Correct")
plt.imshow(imutils.opencv2matplotlib(cactus))
plt.show()
<img src="docs/images/matplotlib.png?raw=true" alt="Matplotlib example"/ style="max-width: 500px;">
This the url_to_image function accepts a single parameter: the url of the image we want to download and convert to a NumPy array in OpenCV format. This function performs the download in-memory. The url_to_image function has been detailed here on the PyImageSearch blog.
url = "http://pyimagesearch.com/static/pyimagesearch_logo_github.png"
logo = imutils.url_to_image(url)
cv2.imshow("URL to Image", logo)
cv2.waitKey(0)
<img src="docs/images/url_to_image.png?raw=true" alt="Matplotlib example"/ style="max-width: 500px;">
The Canny edge detector requires two parameters when performing hysteresis. However, tuning these two parameters to obtain an optimal edge map is non-trivial, especially when working with a dataset of images. Instead, we can use the auto_canny function which uses the median of the grayscale pixel intensities to derive the upper and lower thresholds. You can read more about the auto_canny function here.
gray = cv2.cvtColor(logo, cv2.COLOR_BGR2GRAY)
edgeMap = imutils.auto_canny(gray)
cv2.imshow("Original", logo)
cv2.imshow("Automatic Edge Map", edgeMap)
<img src="docs/images/auto_canny.png?raw=true" alt="Matplotlib example"/ style="max-width: 500px;">
A common task in computer vision and image processing is to perform a 4-point perspective transform of a ROI in an image and obtain a top-down, "birds eye view" of the ROI. The perspective module takes care of this for you. A real-world example of applying a 4-point perspective transform can be bound in this blog on on building a kick-ass mobile document scanner.
See the contents of demos/perspective_transform.py
<img src="docs/images/perspective_transform.png?raw=true" alt="Matplotlib example"/ style="max-width: 500px;">
The contours returned from cv2.findContours are unsorted. By using the contours module the the sort_contours function we can sort a list of contours from left-to-right, right-to-left, top-to-bottom, and bottom-to-top, respectively.
See the contents of demos/sorting_contours.py
<img src="docs/images/sorting_contours.png?raw=true" alt="Matplotlib example"/ style="max-width: 500px;">
The paths sub-module of imutils includes a function to recursively find images based on a root directory.
Assuming we are in the demos directory, let's list the contents of the ../demo_images:
from imutils import paths
for imagePath in paths.list_images("../demo_images"):
print imagePath
../demo_images/bridge.jpg ../demo_images/cactus.jpg ../demo_images/notecard.png ../demo_images/pyimagesearch_logo.jpg ../demo_images/shapes.png ../demo_images/workspace.jpg