2D Real-valued Discrete Gabor Transform and Its Fast Algorithms

A 2D real-valued discrete Gabor transform (RDGT) for image representation is presented in this paper. The Gabor basis functions of the proposed transform are time-shifting windows modulated by cosine waves. Fast algorithms for the computation of the proposed transform coefficients and for the reconstruction of the original image from the coefficients are developed based on 2D discrete cosine transform (DCT). The image coding performance and difference between 2D RDGT and 2D DCT show that the proposed transform is an excellent preprocessing tool for image recognition, image interpretation, texture segmentation, and texture discrimination.