A New Operating Tool for Coding in Lossless Image Compression

1
A New Operating Tool for Coding in Lossless
Image Compression

• Radu Radescu
• University POLITEHNICA of Bucharest,
• Faculty of Electronics, Telecommunications and
  Information Technology
• Department of Applied Electronics and Information
  Engineering
• OC 2009, Varna, Bulgaria

2
1. Introduction

• The application carries out the encoding for
  images with at most as 255 colors.
• As source, the program accepts BMP files,
  represented on 8, 16 or 32 bits.
• The application has an algorithm to reduce the
  color depth to 255 colors.
• As output, the application produces two file
  types, named after the encoding type used to make
  them
• LZW,
• RLC.

3
2. Implementing the compression (a)

• The algorithm
• The colors are stored in a table, which is built
  as the image is run through pixel by pixel.
• If the color of the pixel exists in the table,
  it is ignored.
• If not, it is added to the table.
• Data storage is used, so that the search time
  for the color table can be substantially reduced.

4
2. Implementing the compression (b)

• The encoding flexibility
• In the case of LZW, a variable size dictionary is
  used, with 2048, 4096 or 8192 positions.
• In the case of RLC, there are two ways to run
  through the image
• up and down or
• left to right,
• which exploits in different ways the image
  correlation.

5
2. Implementing the compression (c)

• Reducing the number of colors
• The algorithm is based on the nearest color
  method, computed based on the mean square
  algorithm.
• The generated color palette is a joint one,
  including 128 standard colors, allocated equally
  in the color space.
• Other 127 colors are calculated based on the bar
  graph of the image.

6
3. Experimental results (a)

• The testing files
• A study was made on medical images with different
  degrees of correlation and different number of
  colors.
• Two groups of different-size images were chosen
• one with a large color dispersion,
• other with large areas of the same color.
• Their number of colors was reduces gradually from
  255 to 128 and, finally, to 16 colors.

7
3. Experimental results (b)

• NMR 600600 pixels

8
3. Experimental results (c)

• Angiographies 600600 pixels

9
3. Experimental results (d)

• Tomographies 600600 pixels

10
3. Experimental results (e)

• Ultrasounds 540405 pixels

11
3. Experimental results (f)

• X rays 540405 pixels

12
3. Experimental results (g)

• The LZW and RLC image files are compared to the
  original BMP format and to the archived RAR
  files.
• The compression ratio is superior to GIF because
  of a better saving of coded words in files and of
  the built color palette obtained by the exact
  colors in the image.
• The application has an advantage before the GIF
  compression as the file size is growing and the
  color number is decreasing.

13
3. Experimental results (h)

• The performance of coding for 255 colors

14
3. Experimental results (i)

• The performance of coding for 128 colors

15
4. Conclusions

• The result does not depend much on the number of
  colors from the input image, but on the size of
  the source file.
• The RLC compression depends more on the
  correlation within the image.
• In the case of the 16 color images, the RLE
  compression for the BMP format (with the
  combination of two pixels on the same byte) is
  superior to the RLC format encoding.
• Exceptions are images that, after decreasing the
  number of colors, have become more correlated, in
  advantage of the RLC encoding.

16
Thank you!