Alyn

Documentation coming soon for PyPI version!

Skew detection and correction in images containing text

Image with skew

Image after deskew

###Requires

• numpy
• matplotlib
• scipy
• scikit-image

###Techniques used

• Canny Edge Detection
• Hough Transform

###Features

• Detect the skew in given images
• Display the output
• Save the output to txt file
• Batch process files in a directory
• View Hough Transform of a given image
• Set the number of peaks for Hough Transform and Sigma for Canny Edge detection
• Rotate the image to remove the skew

###How the skew detection works

The skew detection script takes image file as input, then performs the following steps:

• Converts the image to greyscale
• Performs Canny Edge Detection on the Image
• Calculates the Hough Transform values
• Determines the peaks
• Determines the deviation of each peaks from 45 degree angle
• Segregates the detected peaks into bins
• Chooses the probable skew angle using the value in the bins

The deskew script uses the skew angle determined using skew detection script to remove the skew from the image.

###Usage

Image skew calculation using skew_detect.py

To calculate the skew angle for a given image file, use -i option followed by the path to file:

./skew_detect.py -i image.jpg

To save output in a text file add -o option followed by the output file name:

./skew_detect.py -i image.jpg -o output.txt

To display output information add -d option followed by a string Yes:

./skew_detect.py -i image.jpg -d Yes

To batch process files in a directory, use -b option followed by the path to directory:

./skew_detect.py  -b examples

To display Hough Transform plot for an image,:

./skew_detect.py -i image.jpg -p Yes

#####Output of the Hough Transform:

To set the value of sigma for Gaussian blurring in Canny Edge Detection, use -s option followed by the desired value:

./skew_detect.py -i image.jpg -s 3

To set the number of peaks collected from Hough Transform, use -n option followed by the desired value:

./skew_detect.py -i image.jpg -n 10

###Image Deskew using deskew.py

To perform a simple deskew and display the output:

./deskew.py -i image.jpg -d Yes

To save the deskewed image, use the following:

./deskew.py -i image.jpg -o rotated.jpg

In some cases the result image might be upside down or the text may be running vertical, To fix this, use -r followed by the desired angle in int:

./deskew.py -i image.jpg -o rotated.jpg -r 90

To generate data for experimental purposes, run the test_img_gen.py in test_data folder. This will generate images containing a white line having angle between 0 to 180 degrees.