CpE 491: Information Systems Engineering II[1][2]

(Homework 2)

Assigned on: February 12, 2001

Due back on: February 26, 2001

  1. Use the uniform quantizer program provided in the same ftp site to uniformly quantize the Sena image. Do this for various numbers of quantizer bins (N). For each quantized image calculate the distortion due to quantization (D) and plot it against the rate of the quantizer (R ).
  2. Do the exact same thing as above, but also add a Huffman code module at the end. So, now the rate of at which the image is compressed is given by the number of bits you actually spend in encoding the image (after Huffman coding). Is there any difference between the plot that you see in this question and the one you generated in the previous question? Comment on it.
  3. The source code for JPEG is available (for free) at  ftp.uu.net:/graphics/jpeg/jpegsrc.v6b.tar.gz. You can also find the ZIP archive format at ftp.simtel.net:/pub/simtelnet/msdos/graphics/jpegsr6b.zip (for PCs). If you are using “attila” it is already installed there. You can also find these codecs at http://www.ece.stevens-tech.edu/~cpe491/Codecs directory. The command to compress an image is “cjpeg” and the decoder command is “djpeg”. To learn more about how to use these commands you can type “man cjpeg” at the UNIX prompt. Having said all this, here is the question:
    1. Compress the Sena image at various bit rates, starting from 0.75 bits per pixel to about 0.2 bpp. Plot the SNR versus R curve. From the reconstructed images (view them using xview in Solaris or Irfan view or other viewer on PCs) at these rates, what conclusions can you draw regarding the quality of the compressed images? What strikes you most about the distortions you see at the lower rates? Can you comment on what can be done to remove those? Include the PSNR Vs R plot and some images as well in your answer sheet.

 

 

 

 

 



[1] Course web-page: http://www.ece.stevens-tech.edu/~cpe491

[2] K.P. Subbalakshmi, 208 Burchard building