الفهرس 
INTRODUCTIONOnly 14 pages are availabe for public view
 |  | from | 16 |  |  |
| | | from | 16 | | |
Abstract
This thesis consists of seven chapters organized as follows:
Chapter one is an introduction to the field of image and video compression and its applications. At the end, it shows the aim of the present work
Chapter two gives a survey on SIX of the major approaches to image compression which give good reconstructed image quality at different compression ratios (CR):
Subband coding (SBC), wavelet transform (WT), Karhunen-Loeve transform (KL T), discrete cosine transform (OCT), discrete Hartley transform (DHT) and discrete
Gabor transform (DGT).
Chapter three reviews in some details three other image compression techniques and discusses their advantages and drawbacks. These techniques under test are: singular value decomposition (SVD) method, principal component analysis (PCA) and a bit
rate reduction technique for block truncation coding (BTC).
Chapter four briefly describes video compression techniques: different standards, motion compensation, motion detection and temporal subsampling of frames.
Chapter five presents image compression results when applying PCA, SVD ’and BTC methods on four test images. Comparison of the SVD results is provided among the cOlTesponding results of other techniques. A section is devoted here for the results obtained when introducing singular values or principle components of some images to a number of neural networks having the same structure but different weights, for the
sake of image recognition.
Chapter six presents video compression results when applying SVD method on successi ve frames of video sequences. It includes the investigation of two cases of
transmitted frames. Chapter seven contains the conclusions of the thesis and some suggestions for future work; as main conclusions, we can extract the following from the comparison made between different techniques for image compression, based on the quality of reconstructed images in terms of signal to noise ratio (SNR):
a- Whereas DWT and RNN methods give higher SNR values at higher CR, yet the present SVD technique is now competing well with them and better than other methods in this field.
b- SVD method gives better results in the case of images containing edges than those containing details.
c- SVD method is not only better than PCA in image compression but also in image
recognition.
The ap’plication of SVD method on successive frames of video sequences also yields some characteristic results in this field of video compression:
a- SNR values of predi~ted frames are not improved at high CR when increasing the number of transmitted frames .
b- As the compression of the frames increases, the ~ifference between SNR values of predicted and transmitted frames decreases.
c- SNR values of sOl:ne predicted frames remain. ~he same at different CR a~d this could lead to high compression with the same quality.
Finally, it can be seen that SVD is a good technique in the field of image compression, image recognition and video compression.