ayush37 / noisereduction

Demo of implementing CICD pipeline for image noise reduction algos

Noise Reduction Spatial

Demo of implementing CICD pipeline for image noise reduction algos, conducting few sample test cases on CircleCI to test the resulting image reduction

Here I am trying to run few basic Spatial denoise algos over a with 4 differnt types of noises added.

Filters that are implemented:

1. Mean Filter
2. Median Filter
3. Guassian Filter
4. Conservative Filter

First we try to add 4 different type of noise to our image

1. Guassian Noise
2. Poission Noise
3. Sath and Pepper
4. Speckle

python addnoise.py 

This will create 4 distorted images in the images/ directory Now we run our filters on top of these images

To verify the efficacy, we will convert the images to Grayscale and run a noise calculation function before and after applying the filter

python MeanFilter.py 

Output

Mean Filter

##### Running MEAN filter on different noises ####
Estimating quality for guassian image before mean filter   0.6361
Estimating quality for guassian image after mean filter   0.1604

Estimating quality for poisson image before mean filter    0.6416
Estimating quality for poisson image after mean filter    0.1614

Estimating quality for S&P image before mean filter   11.9353
Estimating quality for S&P image after mean filter    0.2798

Estimating quality for speckle image before mean filter   61.4482
Estimating quality for speckle image after mean filter    0.5790

Median Filter

##### Running MEDIAN filter on different noises ####
Estimating quality for guassian image before median filter   0.6361
Estimating quality for guassian image after median filter   0.1981

Estimating quality for poisson image before median filter    0.6416
Estimating quality for poisson image after median filter    0.1995

Estimating quality for S&P image before median filter   11.9353
Estimating quality for S&P image after median filter    0.2358

Estimating quality for speckle image before median filter   61.4482
Estimating quality for speckle image after median filter    1.6339

Guassian Filter

python GuassianBlur.py

##### Running GAUSSIAN filter on different noises ####
Estimating quality for guassian image before Gaussian filter   0.6361
Estimating quality for guassian image after Gaussian filter   0.1825

Estimating quality for poisson image before Gaussian filter    0.6416
Estimating quality for poisson image after Gaussian filter    0.1836

Estimating quality for S&P image before Gaussian filter   11.9353
Estimating quality for S&P image after Gaussian filter    0.2732

Estimating quality for speckle image before Gaussian filter   61.4482
Estimating quality for speckle image after Gaussian filter    0.3154

We can also run UnitTest on these algos

python -m unittest

(DJ) ag841k@US-00010509:~/AG/TopazLabs$ python -m unittest test.py
....
----------------------------------------------------------------------
Ran 4 tests in 13.222s