Joaquin Zepeda
Contact Information
|
Email:
Address: IRISA Campus Universitaire de Beaulieu 35042 Rennes Cedex - FRANCE Phone: +33 2.99.84.25.43 Fax: +33 2.99.84.71.71 Project Assistant: +33 2.99.84.72.28 (Huguette Béchu) Office: F329 |
|
Profile
| (Download resume, July 2010) |  |
| (Download Thesis, October 2010) | |
Nicaraguan national, happily married, and with happy children.
I have successfully defended my Ph.D. on Oct. 28, 2010 and am currently in Technicolor, France, pursuing a post-doc.
- INRIA / Universite de Rennes 1, Rennes, France
- Topic: Image compression and description using sparse representations
Master of Engineering, Electrical Engineering
- McGill University, Montreal, Canada, November 2006
- GPA: 3.77 / 4.00, Dean’s Honor List
- Research group: Multimedia and Signal Processing
Bachelor of Engineering, Electrical Engineering
- McGill University, Montreal, Canada, June 2003
- GPA: 3.53 / 4.00
- Concentration: Telecommunications and Signal Processing
Research Activities
Current topics
- Sparse representations
- Overcomplete dictionary training
- Image compression and processing
- Image description and indexing
An example of an Iteration-Tuned Dictionary (ITD):
ITDs are layered dictionaries with each layer i consisting of a set of dictionaries; one of these dictionaries is used in the i-th iteration of a pursuit decomposition. In the illustration, the dictionary selected is chosen as a function of the atom selected in the previous layer. The colored frames indicate the parent atom of dictionaries of subsequent layers.
Denoising with ITDs:
The PSNR difference plotted is relative to the results obtained using the K-SVD dictionary of Aharon and Elad. The noisy image is obtained using additive white Gaussian noise with variance varying along the abscissa. Results are shown for two ITD variants: the Basic Iteration-Tuned Dictionary (BITD) and the Tree-Structured Iteration-Tuned Dictionary (TSITD).
Compression of image classes using ITDs:
Image compression results for an image codec based on an ITD variant called the Iteration-Tuned and Aligned Dictionary (ITAD). The codec is similar to JPEG but uses ITAD instead of the DCT transform. The rate-distortion curves shown (for various block sizes) are the average over 100 facial images. The sample reconstructed images were encoded using the 12 x 12 blocks and a nominal rate of 0.3 bpp.
Publications
|
J. Zepeda
, C. Guillemot
and E. Kijak Image Compression using Sparse Representations and the Iteration-Tuned and Aligned Dictionary IEEE Transactions on Image Processing (submitted), 2010. |
|
|
|
J. Zepeda
, C. Guillemot
and E. Kijak The Iteration-Tuned and Aligned Dictionary and its Application to Image Compression and Denoising IEEE Transactions on Image Processing (submitted), 2010. |
|
|
|
J. Zepeda
, C. Guillemot
and E. Kijak Iteration-Tuned Dictionaries for Sparse Representations and their Tree-Structured Variant IEEE Transactions on Signal Processing (submitted), 2010. |
|
|
|
C. Morand, J. Benois-Pineau, J.-Ph. Domenger, J. Zepeda
, E. Kijak and Ch. Guillemot
Scalable Object-based Video Retrieval in HD Video Databases Signal Processing: Image Communication, Elsevier (accepted), 2010. |
|
|
|
J. Zepeda
, C. Guillemot
and E. Kijak The Iteration-Tuned Dictionary for Sparse Representations Proc. of the IEEE International Workshop on MMSP, 2010. |
|
|
|
J. Zepeda
, E. Kijak and C. Guillemot
Approximate nearest neighbors using sparse representations Proceedings of the International Conference on Acoustics, Speech, and Signal Processing, 2010. |
|
|
|
J. Zepeda
, E. Kijak and C. Guillemot
SIFT-Based Local Image Description Using Sparse Representations IEEE workshop on Multimedia Signal Processing (MMSP), pp. 1-6, 2009. |
|
|
|
S. Martins, J. Zepeda
, B. Picard, P. Radziszewski and D. Roy Investigating On-The-Shell Acoustics Proceedings of the Autogeneous and Semiautogeneous Grinding Technology Conference, 2006. |
|
|
|
J. Zepeda
and F. Labeau Tandem Filter Bank-DFT Code for Bursty Erasure Correction Proc. IEEE Vehicular Technology Conference, 2006. |
|
|
|
J. Zepeda (85%), C. Guillemot (10%) and E. Kijak (5%) Iteration-Tuned Dictionaries for Sparse Representations (filed) Patent, April, 2010. |
|
| Webmaster: |
|
|
|
| Last time modified: 2011-02-15 |