Multi-focus Image Fusion Using Dual-tree Complex Wavelet Transform

With the development of photography, digital image processing technology are widely used in various fields of applications. Usually with one camera it is difficult to focus on all objects in the same scene. Therefore the technique of image fusion is needed to integrate the information contained in images with different focus points and to get an image with all objects in the scene clearly focused.First the concept, background, and classification of image fusion techniques are introduced. Common methods and objective evaluation criteria are summarized.The Dual-tree wavelet transform is introduced. It inherits the multi-resolution and time-frequency locality feature of the traditional wavelet trasform. Meanwhile it has near shift-invariance, better directionality and limited redundancy. With near shift-invariance, a shift in the input signal won't cause big change in subband energy, which is a problem of traditional wavelet transform. For each decomposion level, 6 directional detail subbands are obtained, extracting image features of±15°,±45°and±75°. Compared to traditional wavelet transform, it can better represent edges and textures in a image.Image fusion methods based on the Dual-tree complex wavelet transform are proposed, in order to solve the problem of poor feature representation with existing discrete wavelet transforms. Multi-resolution representation of the source images is obtained and combined to form the result image. For the high-frequency subbands, a local similarity measure is computed, based on which the fusion rule is selected using either a weighted average or the coefficient with higher local energy. The experimental results are analyzed both subjectively and objectively. The proposed method proves to be superior than methods based on traditional wavelet transform and methods that select the coefficient with higher local energy.