List of Subjects for Project Labs (304-494) or Honours Thesis (304-498 and 304-499)

All the subjects proposed below represent an area in which work can be undertaken in the framework of a project lab (two or more students), and/or an honours thesis (one student).

The depth and extent of the envisioned work will be adapted to the number of students and to the type of project (Lab or thesis).

Most of the projects will involve some theoretical catch-up on the subject, followed by implementation and/or simulations, in which the choice of design methodology and parameters will play a key role.

Of course, if you have an idea of subject and/or area in which you would like to work, drop by to discuss it !

For more information, please contact F. Labeau.

Topics

Image coding and processing

  1. Subjective Video Quality Assessment

     Digital video data is subject to various kinds of distortions during acquisition, compression, processing, transmission and reproduction. It is therefore important to assess and quantify the video quality degradations (in order to control and possibly enhance the quality of the video data). There are two main ways of assessing the quality of digital video:

    • Objective measures, which can be computed mathematically from the original and distorted video signals. 
    • Subjective measures, which consist in quality evaluation by human subjects.
    The most reliable way of assessing the quality of a video is subjective evaluation, because human beings are the ultimate receivers in most applications. The mean opinion score (MOS), which is a subjective quality measurement obtained from a number of human observers, is the most reliable form of quality measurement. However, from a practical point of view the MOS is inconvenient for most applications, as it involves laboratory experiments with human subjects which are time-consuming. Objective measures, on the other hand, are a lot easier to compute as this can be done automatically by a computer.

    Validation of objective measures is an important step. Since the goal of these systems is to predict perceived (subjective) video quality, it is essential to build a video database with subjective evaluation scores associated with each of the video sequences in the database. Such a database can then be used to assess the prediction performance of the objective measurement algorithms.

    The aim of this project is to set up a subjective testing environment and conduct subjective evaluations of digital video data. The key outcome of this project will be a database of subjective evaluations that can then be used to evaluate objective measurements and verify how well they correlate with human perception.

     The project will consist in the following steps: 

    1. Get familiar with digital video standards and the ITU recommendations for subjective quality evaluation of video signals (2 weeks)
    1. Set up a testing environment and protocol; Select the test sequences. (2 weeks)
    1. Recruit volunteer test subjects and carry out test sessions. (4 weeks)
    1. Gather the results into a database; perform some statistical analysis on the data (calculation of mean scores, confidence intervals, etc…) (1 week)
    1. Use the subjective results in the database to evaluate the prediction performance of existing objective measures; (3 weeks)
    1. Publish a report

  2. Design and implementation of a Network Video Transmission system

    Co-supervisors: Prof. Labeau and Prof. Le-Ngoc

    Background:   The main topic of this project is the design and implementation of a video transmission system of packet-oriented networks. The application (see figure below) considered is the case where a video source is fed to users through an IP based network (such as the internet). The video needs to be compressed in order to accommodate the limited capacity of transmission of the network.

     

    Objectives: The aim of this project is to design and implement a video compression system (based on existing standards, namely MPEG-4 and H.264 [1]) that will be able to adapt the compression ratio (through Rate control)to changing network conditions: when the network is heavily loaded, the compression ratio needs to be increased so as to decrease the number of packets to be transmitted; on the other hand, when the network conditions are favorable, the compression ratio can be decreased to allow for better video quality. In order to design this video compression system, the following points will have to be addressed:

    • Get familiar with existing video compression standards (MPEG-4 and H.264) and their software simulators
    • Setup a software simulation testbed for the system, comprising video source, compression, networks simulation, rate control strategy and reproduction.
    • Devise and test rate control strategies to adapt to varying network conditions; the factors to take into account and figures of merit will be: overall video quality, speed of adaptation, and subjective (visual) quality.
    • Finally, in parallel, the system will be implemented on a hardware simulator (based on DSP boards and PCs).

    Supervision: Work will be carried out in the Multimedia signal Processing Laboratory, under the supervision of Profs Labeau and Le-Ngoc, and in close collaboration with a PhD student working on the topic.

    Pre-requisites and qualifications:

    Due to the extent of the work and the fact that it is closely related to ongoing research projects, interested students should be motivated and ready to work hard during the whole semester; a group of more than 2 students or 2 groups of 2 students would be welcome.

     It is desirable that students willing to undertake this project have had some prior exposition to signals and systems (ECSE-303 and 304 REQUIRED).

    Ideally, students should have had some prior exposition to embedded systems (ECSE-426, ECSE-422 or ECSE-436), as well communications networks (ECSE-414).

    Reference:

    [1]     G.J. Sullivan and T. Wiegand: Video Compression - from concepts to the H.264/AVC standard, Proceedings of the IEEE, Vol. 93,  No. 1,  Jan. 2005, pp. 18 – 31. (Available on IEEEXplore)

  3. Error correction on image multiple description transmission.

    4.

Multiple Description Coding is a technique through which several descriptions or approximations of the same signal are sent over a network incurring possible packet losses and errors. At the receiving end, if all descriptions are received, a good reproduction of eh original signal is possible, whereas, if only one of two descriptions is received, a rougher approximation is made for reproduction.

In our recent research, we have developed a novel technique to reconstruct lost portions of descriptions when only one of them is received, by using the residual redundancy between the two descriptions. This technique has only been implemented up to now on abstract and theoretical data.

The aim of this project is to implement our technique in a full-blown image compression system, using standard compression techniques and multiple description coding.

Reference:

[1]    V. K. Goyal, Multiple description coding: compression meets the network, IEEE Signal Processing Magazine, Vol 18, No 5, Sept. 2001, pp.74-93 (Available on IEEEXplore)

  • Multiple Description Video Transmission

    •

    Multiple Description Coding is a technique through which several descriptions or approximations of the same signal are sent over a network incurring possible packet losses and errors. At the receiving end, if all descriptions are received, a good reproduction of eh original signal is possible, whereas, if only one of two descriptions is received, a rougher approximation is made for reproduction.

    The aim of this project is to implement a multiple description coding video transmission technique based on standard coders, and using very simple techniques for production of the two descriptions (techniques such as time decimation).

    Reference:

    [1]    V. K. Goyal, Multiple description coding: compression meets the network, IEEE Signal Processing Magazine, Vol 18, No 5, Sept. 2001, pp.74-93 (Available on IEEEXplore)

     

    Signal Processing

    1. Design and Implementation of applications for DSP education:

      The aim of this project is to design and implement some MATLAB applications (animations/graphics/GUIs) to be used in DSP courses by the instructor to demonstrate DSP related concepts. Example of possible applications are: illustration of phase/amplitude distortion and filtering on speech and audio signals, SIMULINK models to illustrate the effect of pole/zero placement on filter responses, 3-D representation of Z-transform surfaces and link to DTFT,...

      The work will involve close collaboration with the course instructor in order to understand the specifications and requirements of each demo.

         Students who choose this project should already have taken ECSE-412 or ECSE-512, and passed it with a high grade. A good knowledge of MATLAB is of course an asset.

       

    2. DFT based error correction:

      Pre-requisites: Discrete Time Signal Processing (ECSE 412)

      The effects of channel distortions on digital data transmission systems can be measured as a percentage of erroneous bits (bit error rate) at the receiver. Under practical conditions the bit error rate of raw transmitted data can be too large for useful applications. In order to combat channel errors and bring down the bit error rate, redundancy is added to the raw transmitted data using an arsenal of different mathematical constructs known as channel codes. Parity bits are a very simple example of a channel code.

      Digital data transmitted over networks is commonly organized into packets of data. Due to channel imperfections and network traffic, these packets can become corrupt during the transmission or simply become lost altogether, resulting in erroneous and "erased" samples.

      Recent work has been done on the usefulness of a special type of code known as the DFT code in combating channel errors as well as reconstructing channel erasures. DFT codes work on real-valued (or complex-valued) discrete data (note at time n, the discrete signal x[n] takes values over the entire complex plane). They consist of a matrix multiplication mapping a real-valued K-dimensional vector to a real valued N-dimensional vector, N>K. The mapping allows the detection and exact correction of up to floor[(N-K)/2] errors. The code can also exactly reconstruct up to N-K erasures.

      This project overlooks erroneous samples and addresses the case of erasure reconstruction. Since DFT-codes operate in the real-field, quantization associated with digital transmission results in inexact erased sample reconstruction. In particular, when the erasures are adjacent across entries of the transmitted vector (bursty erasures), the reconstruction error is very large. A code based on the DFT-Code is being developed at McGill as part of a Master's thesis. Rather than mapping K-D vectors to N-D vectors, the code maps the input data to matrices. The reconstruction can then be carried out along the rows and columns of this matrix, where the reconstruction order is selected in order to break up bursty erasures.

       The current project would consist of the following 2 tasks:

      1) Implementing Linux C code to simulate the encoding, transmission, data loss, and data reconstruction as a function of the reconstruction order.

      2) Trying various decoding orders in order to explore how well the decoding order affects the mean square error of the reconstructed vectors.

       

    Other

    1. Development of a Microsoft  Excel® template for automated course marks management

    This is a software engineering oriented project. The aim of the project would be to provide professors with a standardized template spreadsheet for entering student marks for different courses. 

    The project would include a period of specification of needs with the users, as well as the specification of interfaces with different existing mark management schemes, student listings from the university and the electronic mark submission system. 

    This project requires the knowledge of Visual Basic.

     

    (*) Note that you will only have access to the hyperlinked papers when connecting from McGill or through VPN

    [TSP lab] [ECE department] [McGill University]
     
    Maintained by F. Labeau. Last Modified January 12, 2006 .