Nonparametric regression modeling with topographic maps as a basis for lossy image compression

Author

Van Hulle, Marc M.

Author_Institution

Lab. voor Neuro- en Psychofysiologie, Katholieke Univ., Leuven

fYear

1997

fDate

24-26 Sep 1997

Firstpage

4

Lastpage

13

Abstract

We introduce a new approach to lossy image compression with topographic maps, a type of neural net, based on nonparametric regression modeling: the topographic maps are trained to perform nonparametric regression using the author´s maximum entropy learning rule (1995, 1997), in combination with projection pursuit regression learning. Furthermore, in order to better account for the local image statistics, we apply a technique similar to subspace classification. Finally, we compare the performance of our approach to that of the Karhunen-Loeve transform and the optimally integrated adaptive learning algorithm

Keywords

data compression; image coding; learning (artificial intelligence); maximum entropy methods; neural nets; nonparametric statistics; statistical analysis; Karhunen-Loeve transform; local image statistics; lossy image compression; maximum entropy learning rule; neural net; nonparametric regression modeling; optimally integrated adaptive learning algorithm; projection pursuit regression learning; subspace classification; topographic maps; Clustering algorithms; Entropy; Image coding; Karhunen-Loeve transforms; Laboratories; Neural networks; Neurons; Partitioning algorithms; Statistical distributions; Vectors;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks for Signal Processing [1997] VII. Proceedings of the 1997 IEEE Workshop

Conference_Location

Amelia Island, FL

ISSN

1089-3555

Print_ISBN

0-7803-4256-9

Type

conf

DOI

10.1109/NNSP.1997.622378

Filename

622378