Welcome to my web page. I started my PhD work in the year 2001, with the title “mathematical tools for image treatment and image perception”, since that date I had the chance to proceed in research concerning Image processing with Application to Image Filtering and Image compression and perception. Technically speaking my research efforts are concentrated in the following domains:
1-2D Positive definite Toeplitz Systems: Studying and enhancing a wide spread families of Toeplitz matrix inversion, and linear prediction Algorithms, my main objective is to fully adapt Multi-channel Algorithms (WWR, 2D Burg Algorithm) to the 2D case, this of course did lead us to better understanding of the nature of problem and the development of new algorithms. One of the main application areas is Adaptive Signal Processing for Radar Signals, SAR, 3D and 3D magnetic resonance medical imaging IMR, and faster Image compression and decompression algorithm.
2-Vectorial Image Quantization: Image Quantization is a well-known method for image and signal compression. In this method the image compression is done first by segmenting the image into a grid of blocks, and then followed by an optimization process to replace this group of blocks with another set of blocks with lesser number of elements. To our advantage 2D linear prediction is also one of the methods used in Image quantification, and my objective is to investigate the different methods already present and to propose new ways to accelerate their implementation.
3-Object Perception: In connection to image compression and object identification, Object Perception is a new concept that I tried to develop during the period of my thesis. Object Identification is a very important field, which had a lot of attention from researchers from all over the world. The new in Object Perception concept is the independence of the identification process from any A Priori information. This leads us to its main characteristic as an autonomous process and that is the adaptability to multiple object definition and eventually being able to provide a reasonable Projection toward signal development and signal interpretation.
Multidimensional Linear Prediction
Rami
KANHOUCHE
Linear Prediction is far from being the most intelligent predictor; still it is very useful in cases where the signal got a stationary nature. In other words, it got almost a repetitive structure. In the picture above we see the result of applying a 10 by 10, 2D linear predictor filter over the exterior edge of the image. In linear prediction a prediction filter is learned over the signal to be used after in different ways:
| · |
Signal Spectral Estimation |
| · |
Signal de-noising. |
| · |
Texture Signal Classification. |
| · |
Signal development (as in the example above). |
Copyright © 2007 Golansoft · All Rights reserved
Last Updated: 1-2-2007
Rami kanhouche, PhD.
App. 409, 35 rue du docteur babinski
Paris 75018, France.
Tel: 00 (33) 6 62 29 82 19
