(Semi) Automated Image Processing with pyautocv

Project Aims

The goal of pyautocv is to provide a simple computer vision(cv) workflow that enables one to automate or at least reduce the time spent in image (pre)-processing.

Installing the package

From PyPI

pip install pyautocv

From GitHub

pip install git+https://github.com/Nelson-Gon/pyautocv.git
#or
# clone the repo
git clone https://www.github.com/Nelson-Gon/pyautocv.git
cd pyautocv
python3 setup.py install

Example Usage

Note: Although these methods can be run via this script, the script is less flexible and might be useful for quick exploration but not extended analysis.

To run the script at the commandline, we can do the following

python -m pyautocv -d "images/cats" -s "png" -m "thresh_to_zero" -o "threshold" -mt 200 -t 100

Sample Result

To perform edge detection

python -m pyautocv -d "images/biology" -s "jpg" -o "detect_edges" -m "sobel_vertical" -k 3

To smooth images

python -m pyautocv -d "images/houses" -s "jpg" -o "smooth" -m "gaussian" -k 5 5 --sigma 0.7

To get help

python -m pyautocv -h 

Further exploration is left to the user.

---

The following section shows how to use the more flexible class/methods approach

• Image Gra(e)ying

To grey an image directory

from pyautocv.segmentation import Segmentation, gray_images, show_images

images_list=Segmentation("images/cats")
show_images(gray_images(images_list.read_images()), images_list.read_images(), number=2)

• Smoothing

To smooth a directory of images, we can use EdgeDetection's smooth method as follows

images_list=Segmentation("images/cats")
show_images(images_list.smooth(), images_list.read_images(),number=2)

This will give us

The above uses default parameters including an rgb color mode. For biological images which are often in grayscale, one can set color_mode to gray as shown below. All other operations will remain the same.

images_list_gray_mode=Segmentation("images/dic", image_suffix ="tif", color_mode = "gray")
show_images(images_list_gray_mode.read_images(), images_list_gray_mode.threshold_images(), number = 4)

Result

To use a different filter

images_list = Segmentation("images/cats")
show_images(images_list.read_images(), images_list.smooth(mask="median", kernel_shape=(7, 7)))

• Edge Detection

To detect edges in a directory of images, we can use Segmentation's detect_edges.

show_images(images_list.read_images(), images_list.detect_edges(operator="roberts", mask="gaussian", sigma=0.8))

The above will give us the following result

To use a different filter e.g Laplace,

show_images(images_list.read_images(), images_list.detect_edges(operator="laplace", mask="gaussian", sigma=0))

This results in

• Thresholding

To perform thresholding, we can use the method threshold_images.

to_threshold = Segmentation("images/biology")
show_images(to_threshold.read_images(),to_threshold.threshold_images())

To use a different thresholding method.

show_images(to_threshold.read_images(),to_threshold.threshold_images(threshold_method="otsu"))

The above gives us:

For cat lovers, here's thresholding with inverse binary.

show_images(images_list.read_images(),images_list.threshold_images(threshold_method="binary_inverse"))

Result:

Thresholding applied to images of houses.

images_list=Segmentation("images/houses")
show_images(images_list.read_images(), images_list.threshold_images(threshold_method="thresh_to_zero"))

images_list=Segmentation("images/potholes")
show_images(images_list.read_images(), images_list.threshold_images("binary"))

These and more examples are available in example.py. Image sources are shown in sources.md. If you feel attribution was not made, please file an issue and cite the violating image.

Citation

Nelson Gonzabato(2021) pyautocv: (Semi) Automated Image Processing, https://github.com/Nelson-Gon/pyautocv.

@misc {Gonzabato2021,
author = {Gonzabato, N},
title = {pyautocv: (Semi) Automated Image Processing},
year = {2021},
publisher = {GitHub},
journal = {GitHub repository},
howpublished = {\url{https://github.com/Nelson-Gon/pyautocv}},
commit = {7fe2e0f7894e8be4588a22758e8097c247cd1cd9}

Thank you very much

“A language that doesn't affect the way you think about programming is not worth knowing.” ― Alan J. Perlis

---

References

• Bebis

• Standford, author unknown

• Funkhouser et al.,2013