Abstract: According to the different frequency areas decomposed by Curvelet transform, the selection principles of the low frequency coefficients and the high frequency coefficients were discussed respectively. In choosing the low frequency coefficient, the concept of the local area standard deviation was defined and a scheme combining the “selection” and the “average” methods was employed. In choosing the high frequency coefficient,taking the advantage of the directional sensitivity characteristic of the Curvelet transform, the concept of directional contrast in the Curvelet domain was defined and a selection principle based on the direction contrast was presented. The experimental results show that the proposed algorithm can extract all useful information from the original images and make all targets in the fused images very clear.

Key words: information processing, image fusion, Curvelet transform, directional contrast, local area standard deviation