Paul van Diepen - Full Color Spatial Frequency Filtering of Scenes

Scene Perception Research, October 1992 - April 1999

Laboratory for Experimental Psychology
University of Leuven, Belgium

Back to the index page

Introduction

There is no straight-forward way to apply Fourier filtering to full-color images. We chose the following approach: First, the full-color image is decomposed into RGB components. Each of these components is then filtered separately, as if it way an ordinary grey-level image. Finally, the three components are joined again. We used Matlab Version 4.2c.1 to perform the image processing. The source code of a Matlab demonstration program is available on this site. Below a computer-generated image is shown, together with low-, band-, and high-pass filtered versions of it. In our experiments, images subtended 16 x 12 visual degrees. Low-pass and high-pass cut-off frequencies were 1.0 cpd and 3.0 cpd respectively.

Unfiltered Stimulus

original stimulus

Low-pass filtered stimulus

low-pass filtered stimulus

Band-pass filtered stimulus

high-pass filtered

High-pass filtered stimulus

high-pass filtered
Paul M. J. van Diepen - October 2002