|
自动化学报 2007
Fast Gabor Filtering
|
Abstract:
In this paper,two fast implementations of Gabor filter are proposed.The implementation strategy involves two steps:1)decomposing Gabor filter into 1-D Ganssian filters along non-orthogonal axes with different variances; 2)recursively implementing these 1-D Gaussian filters.The non-orthogonal decomposition and recursive filtering ensure efficient performance.Experimental results showed that the proposed implementations outperform traditional convolution filtering with respect to computational speed and accuracy.