The development

1
The development integration of a Real-Time
X-Ray image transfer system over a wireless
networkFinal Year Project PresentationCapt
David Clarke
2
Outline Of Presentation

• Background to Project
• Technology Research
• Project Overview
• Image Compression
• Matlab
• MSE
• Gumstix
• Main tasks Completed

3
Background to Project

• The Army Ordnance Corps recently upgraded their
  HOBO robot
• Using COFDM for video link
• Also use a manual x-ray system
• Reamda Ltd investigating use of
• COFDM or ethernet
• to send x-ray picture wirelessly
• Intention was to investigate the sending of a
  compressed image
• Investigate image comparison

4
(No Transcript)
5
(No Transcript)
6
Technology Research

• COFDM Coded Orthogonal Frequency Division
  Multiplexing
• Non-Destructive Testing Digital X-Ray
• TIF and JPEG
• Gumstix and Linux

7
COFDM

• Multi-carrier transmission technique
• Uses bandwidth efficiently
• Very robust against channel interference
• Uses a Guard Interval

8
Non-Destructive Testing

• Involves Digital Radiography
• Gives Instant read-out of an x-ray image
• Currently it is Manually Deployed
• Images are quite large
• Must be compressed in order to transmit
• Non-Destructive testing is used in many
  industries

9
JPEG

• Standardised image compression algorithm
• Lossy image
• Trade-off of file size vs image quality
• Mathematical formula
• Source code freely available
• Many myths about JPEG

10
Project Overview

• Research
• Sample x-ray images compressed
• from TIF to JPEG
• Image Quality assessed by experts
• and using Matlab
• MSE method investigated
• Gumstix used to send new images

11
Image Compression

• Simply done using ImageConverter PLUS
• TIF files of 13 MB reduced to100 300KB
• when converted to JPEG
• On using Gumstix, it was found that zipping the
  files reduced size to approx 400 600KB
• 100 Quality
  85 Quality

12
Image Compression Assessed

• Compressed Images sent to Ordnance Workshops and
  assessed by experts
• All 6 images sent were approved for use (100
  Quality)
• 5 out of 6 images were approved for use (85
  Quality)
• The images were scored out of ten Mean and SD
  calculated
• JPEG Myth ?

13
Matlab

• Matlab used to find the MSE Mean Square Error
  of before and after images
• Using MSE, PSNR Peak Signal-to-Noise Ratio was
  then calculated
• PSNR 10log10((MAX(L))2 / MSE))
• Plotted against Compression rate

14
MSE

• Mean Square Error has many good attributes
• Simple to use Natural to define energy of a
  signal Possesses symmetry and
  differentiability - Widely used in signal
  processing
• However, using it to predict human perception of
  image quality is not deemed a good idea
• Too many assumptions Random re-ordering
• - MSE wont pick up an error
• - Signal fidelity as important as signal
  error samples

15
Gumstix

• Connex 400xm
• Used to send the image to the host PC
• Linux had to be learned
• http//ohm.nuigalway.ie/wiki
• http//docwiki.gumstix.org
• Simple commands used to zip image and scp to host
  pc

16
Main Tasks Completed

• Research of Technology
• Knowledge gained of Army Equipment
• Image Compression and Comparison
• Matlab
• Understanding of Linux
• First time to use Gumstix

17

• Thank you for your attention