Abstract:

We propose a new image fusion algorithm based on finite ridgelet transform and cycle spinning. The finite ridgelet transform provides a flexible multiresolution, local and directional image expansion, and also a sparse representation for two-dimensional piecewise smooth signals resembling images. However, finite ridgelet transform lacks translation invariance that the fused image introduces many artifacts. We use the cycle spinning technique to develop translation invariant contourlet fusion rule. This method can effectively eliminate the Gibbs phenomena, extract the character of original images and preserve more important information. Subjective and objective standard evaluations of the experiment results show that this proposed method outperforms conventional methods in terms of better image character extraction and edge character description.

Key words: computer application, image fusion, finite ridgelet transform, cycle spinning, Gibbs phenomenon