Use fuzzy logic to implement a selective median filter to clean noisy images

1
Use fuzzy logic to implement a selective median
filter to clean noisy images

• HUI LI
• BSU
• 11/14/2000

2
Importance

• The process of image acquisition frequently
  leads(inadvertently)to image degradation .
• Develop a method to solve this program.
• Literature review

3
Research step

• Learn Win32 API programming
• Learn fuzzy logic basics
• Learn how to load bitmap using VC
• Produce test noise on image
• Algorithm to filter image

4
What is fuzzy logic

• Ordinary(Crisp) Sets and Fuzzy Sets
• Use fuzzy logic to deal with uncertainty
• Construct membership function and rules

5
Bitmap format

• Bitmap-file header contains type,size,layout of
  a device-independent bitmap file
• Bitmap-information header contains the
  dimensions, compression type, color format
• Color table (RGB)

6
What is median filter

• Bitmap files formats used by the Microsoft
  Windows
• Defines a neighborhood of the pixel

7
Bitmap noise
8
Continue
9
About load images

• Get pixel value and set to the context

10
Produce test noise

• Produce random value(0-255) to replace the
  original pixel value(random point)

11
How to use fuzzy logic to my research paper

• Construct membership functions and rules
• Fact 1,difference between highest 5 pixel values
  and 5 lowest pixel values
• Fact 2,difference between considered point pixel
  value to average value(55 point pixel values)

12
Data analysis

• Determine which point should be filtered
•  fact 1
• 1
• 0 low med large 255

13
continue

•  fact 2
• 1
• 0 low med large 255

14
Continue

• small median
  large
• Small not replace replace
• Median replace not replace
• Large replace replace not

15
Schedule

• Learning Windows programming need more time
• The total time need 7-8 months

16
Facilities and Special Resource

• Hardware ,Operating System,VC environment

17
Deliverables

• This research can be finished using above
  computer program

18
Reference

• Inside Visual C (David J.Kruglinski)
• Windows 95 with MFC (Jeff Prosise)

19
THANK YOU