Neural networks for fast image compression /

algorithm was adopted. In order to keep the

Bibliographic Details
Main Author: Li, Mu
Format: Thesis eBook
Language:English
Published: [Place of publication not identified] : [publisher not identified] ; 1998.
Subjects:
Online Access:Link to OAKTrust copy
Description
Summary:algorithm was adopted. In order to keep the
and decrease the number of neurons in the intermediate
aspects: compression ratio, image distortion, and
based on the predictive values and the input vectors.
compress a gray-scale image. The system consists of
compression part includes the input layer and the
compression ratio. Moreover, the parallel architecture
consists of the intermediate layer and the output
converge, an adaptive back-propagation learning
design of an image compression system involves three
fidelity of the image, such that at the receiver, the
for the learning process of the neural networks to
generalization capability of the compression system
good Peak-to-peak Signal to Noise Ratio and high
image, a set of natural networks instead of one
implement a new compression/decompression system to
inherent in the architecture of the neural networks
intermediate layer, while the decompression part
layer, a preprocessing element is designed which
layer. To gain high quality of the reconstructed
makes the compression system process very quickly.
network have been used in the system. Each neural
network was trained with some image blocks which have
number of bits transmitted as well as keeping the
performs the necessary' processing of the image before
predictor which predicts the current input block and a
processing speed. In this thesis, we design and
reconstructed image will have little distortion. The
similar characteristics. In order to decrease the time
subtracting element which generates residual vectors
The final results shows that the reconstructed image
The image compression system aims at reducing the graphics.
the image is encoded. The preprocessor includes a
three-layer feed forward neural networks. The
which was processed by our proposed scheme had a very
Item Description:"Major subject: Electrical Engineering".
Vita.
Physical Description:x, 62 leaves : illustrations ; 28 cm.
Also available online.
Bibliography:Includes bibliographical references (leaves 42-44).