Scribd
Upload
69 views8 pages

Es 2 Final Project Powerpoint

This document summarizes an image correction project that aims to remove pixel-by-pixel random noise from corrupted images. It explores two approaches: non-selective median filtering and selective filtering, where the median is only applied if the difference between pixels exceeds a threshold. Results show that selective filtering effectively mitigates distortion at different noise levels. However, it struggles on busier images where large differences may be due to the subject rather than noise. The next steps are to improve handling of busier images and explore machine learning approaches.

Uploaded by

Andrew Sack
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
69 views8 pages

Es 2 Final Project Powerpoint

This document summarizes an image correction project that aims to remove pixel-by-pixel random noise from corrupted images. It explores two approaches: non-selective median filtering and selective filtering, where the median is only applied if the difference between pixels exceeds a threshold. Results show that selective filtering effectively mitigates distortion at different noise levels. However, it struggles on busier images where large differences may be due to the subject rather than noise. The next steps are to improve handling of busier images and explore machine learning approaches.

Uploaded by

Andrew Sack
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

Image Correction

Andrew Sack
May 12, 2017
Abstract:
A common issue when collecting and using data is data loss or
corruption. This project focuses on the issue of corruption in image
data, specifically in the form of pixel-by-pixel random noise It aims
to approximate the non-corrupted image using just the corrupted
image. Using statistical analysis of the images and variations on
median filtering, this program can identify and mitigate or remove
the corrupted pixels.
Background
• A problem with all things involving data is
collecting good data
• When that is not possible, it is important to
be able to handle bad data; being able to
identify and correct the bad data is critical.
• Specifically in images, it is common to have
noisy or corrupted
•This project focuses on corruption in the form
of pixelated noise
Technical Approach – 1
Non-Selective Filtering
• Uses modified method of 2D median filtering
• The function steps through each point on
each channel in the image.
•For each pixel, the value of each pixel in the M
x N box around the pixel (excluding the target
pixel) is added to an array.
• The median of this array is taken and that
becomes the value of that pixel in the
corrected image.
Technical Approach – 2
Selective Filtering
• For selective filtering ONLY, the absolute value
of the difference between the center pixel and
each pixel in the M x N box surrounding it is
added to an array.
•The median is taken, and only if the median is
greater than a threshold median difference
value, the value of the pixel is corrected to the
median.
Results- 1
Efficacy of Correction on Different Levels of Distortion
(all with default parameters and selective correction)

20% 40% 60% 80% 100%


Before

After
Results- 2 Found Pre-Distorted

3 Iterations of
Selective Filtering
with default
parameters

2 Iterations of
Selective Filtering
with default
parameters
Conclusions and Next Steps
• Determine a way to handle busier images
better
–Current method does can not distinguish large
differences due to noise from large differences
due to the subject of the image
• Look into a machine learning approach to
solve this.
•Improve speed and efficiency of code
References
1. Miller, Eric (2009). Region Processing II
[PowerPoint slides]. Retrieved from
trunk.tufts.edu

You might also like