Using reinforcement learning for filter fusion in image enhancement

Farhang Sahba, Hamid R. Tizhoosh, Magdy M.A. Salama

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

A new approach to image enhancement based on fusion of filters using a reinforcement learning scheme is presented. In most applications the result of applying a single filter is usually unsatisfactory. Appropriate fusion of the results of several different filters, such as median, local average, and Wiener filters can resolve this difficulty. Many different techniques already exist in literatures. In this work, a reinforcement learning agent will be proposed. We use this novel method as an effective way to find appropriate weights for various filters. During learning, the agent takes some actions (i.e., different weights for filters) to change the environment (the image quality). Using the reference image, the RL agent is provided by a scalar evaluation determined objectively. The agent uses these objective rewards-punishments to explore/exploit the solution space. The values obtained using this method can be used as very valuable knowledge to initiate weights for a Q-matrix. A subjective reinforcement learning method can use such initial weights to start its work from an acceptable level of knowledge.

Original languageEnglish (US)
Title of host publicationProceedings of the IASTED International Conference on Computational Intelligence
Pages262-266
Number of pages5
StatePublished - 2005
EventIASTED International Conference on Computational Intelligence - Calgary, AB, Canada
Duration: Jul 4 2005Jul 6 2005

Publication series

NameProceedings of the IASTED International Conference on Computational Intelligence
Volume2005

Conference

ConferenceIASTED International Conference on Computational Intelligence
Country/TerritoryCanada
CityCalgary, AB
Period7/4/057/6/05

Keywords

  • Filter fusion
  • Image processing
  • Machine learning
  • Reinforcement learning

ASJC Scopus subject areas

  • Engineering(all)

Fingerprint

Dive into the research topics of 'Using reinforcement learning for filter fusion in image enhancement'. Together they form a unique fingerprint.

Cite this