Reference no: EM132511981

CIS115-6 Signals and Electronic Systems Assignment - University of Bedfordshire - Majan College, Oman

Aim -

Gain experience of programming and signal processing algorithm implementation in Matlab.

Develop team work and presentation skills.

Task - You are designing an image compression system as a component of a new video monitoring system for a robotic manufacturing line. Your task is to optimize the image compression system so that the reconstructed decompressed images have PSNR of at least 31dB achieved with the minimal possible number of retained transform coefficients.

Assume the image compression system uses the wavelet transform with Daubechies-2 filter-bank. Then, run compression iteratively for the number of decomposition levels J changing from 1 to 9 and apply this to a selection of at least 5 images from the test image folder. Then, for each case, identify the number of retained (non-zero) transform coefficients for which the system achieves the PSNR of 31dB. Finally, by comparing the obtained performance, select the best number of decomposition levels J such that the identified number of retained coefficients is minimal.

For clarity, this list of tasks is tabularized below:

A. Select J from 1 to 9 and apply the forward wavelet transform using the Daubechies-2 filter-bank.

B. Select a threshold for the magnitude (absolute value) of transform coefficients and apply it - zero out all the transform coefficients that have the magnitudes below this threshold.

C. Apply the inverse wavelet transform using the same number of decomposition levels J and the same filter type.

D. Compare the reconstructed image to the original one using the PSNR calculation.

E. If PSNR is smaller than 31dB, return to B using a smaller threshold (the image is too compressed and you need to discard less information); if PSNR is larger than 31dB, return to B using a larger threshold (the image is too good and we can afford discarding more information in order to reduce the number of retained coefficients).

F. Once you find a threshold that achieves PSNR of 31dB (plus / minus 0.2db), denote the number of retained coefficients as the minimal rate for that case.

G. Repeat A-F for different J until you find out the minimal rate for each J between 1 and 9.

H. Compare the minimal rates and select J that results in the smallest minimal rate.

Write an m-script that would implement the steps A-H above. Make sure you save the graphs of plotted PSNR values against the percentage of retained coefficients as evidence of your findings for each tested image and for all values of J plotted on the same graph using different colours or line style.

Then write a report that would explain the process how you analysed the problem and how you found out the solution (the best configuration of the wavelet transform, i.e. selected J).

While accomplishing the task, you are allowed to use the Matlab source code provided in the practical sessions or any other available Matlab code on Internet (with proper referencing). You are also allowed to form teams of 2 or 3 members to accomplish the task. The presentation files have to give clearly the names of the team members on the first page.

Upon completing the software development and image approximation evaluation task, you have to submit the team presentation and individual reports to Breo. Templates with suggested content for both presentation and report will be made available via Breo.