Rate constrained self-organzing neural maps and efficient psychovisual methods for low bit rate video coding

Ferguson, K.L. and Allinson, N.M. (1999) Rate constrained self-organzing neural maps and efficient psychovisual methods for low bit rate video coding. In: Neural Networks for Signal Processing IX, 1999, 23-25 August 1999, Madison, WI, USA.

Documents
Rate-constrained self-organizing neural maps and efficient psychovisual methods for low bit rate video coding
[img]
[Download]
Request a copy
[img] PDF
00788158.pdf - Whole Document
Restricted to Repository staff only

530kB

Full text URL: http://ieeexplore.ieee.org/xpl/articleDetails.jsp?...

Abstract

The video coding problem is essentially an operational distortion-rate issue where the underlying input pixel data, probability distributions and dimensions are discrete, unknown and not smooth. In the low bit rate case the high resolution assumptions for vector quantization are not strictly valid and the problem is exacerbated. However, by considering the rate-constrained operational points on sets of self-organizing neural maps (SOMs), provides a methodology for selecting locally optimal vector quantizers. The learning process of the standard SOM algorithm is modified to minimize the distortion subject to a constraint of entropy approximation. The applied training set is adapted to suit the proposed coding environment. Operating in the discrete wavelet transform (DWT) domain is well suited to the inclusion of a psychovisual model. The spatial frequency response, the multiresolution scene analysis and the central focusing aspects of the visual cortex are incorporated into the model. The resulting video coding algorithm is bit rate scalable from 10 k bits per second (bits/s) and provides subjectively acceptable video at a fixed frame rate or 10 frames per second (f.p.s.) with a QCIF pixel resolution

Item Type:Conference or Workshop Item (Paper)
Additional Information:The video coding problem is essentially an operational distortion-rate issue where the underlying input pixel data, probability distributions and dimensions are discrete, unknown and not smooth. In the low bit rate case the high resolution assumptions for vector quantization are not strictly valid and the problem is exacerbated. However, by considering the rate-constrained operational points on sets of self-organizing neural maps (SOMs), provides a methodology for selecting locally optimal vector quantizers. The learning process of the standard SOM algorithm is modified to minimize the distortion subject to a constraint of entropy approximation. The applied training set is adapted to suit the proposed coding environment. Operating in the discrete wavelet transform (DWT) domain is well suited to the inclusion of a psychovisual model. The spatial frequency response, the multiresolution scene analysis and the central focusing aspects of the visual cortex are incorporated into the model. The resulting video coding algorithm is bit rate scalable from 10 k bits per second (bits/s) and provides subjectively acceptable video at a fixed frame rate or 10 frames per second (f.p.s.) with a QCIF pixel resolution
Keywords:discrete wavelet transforms, entropy, frequency response, image resolution, probability, rate distortion theory, self-organising feature maps, teleconferencing, transform coding, vector quantisation, video coding, visual perception
Subjects:G Mathematical and Computer Sciences > G730 Neural Computing
Divisions:College of Science > School of Computer Science
ID Code:5128
Deposited By: Tammie Farley
Deposited On:02 May 2012 06:23
Last Modified:13 Mar 2013 09:06

Repository Staff Only: item control page