Union Laplacian pyramid with multiple features for medical image fusion

• Affine transformation is introduced in pyramid achieving multi-orientations.
• Kirsch method is used to highlight the contrast of the images.
• PCA method is for highlighting the contrast of the images.
• The averaging different orientation is to preserve the structure.
• Histogram of images in experimental part is to evaluate contrast.

The Laplacian pyramid has been widely used for decomposing images into multiple scales. However, the Laplacian pyramid is believed as being unable to represent outline and contrast of the images well. To tackle these tasks, an approach union Laplacian pyramid with multiple features is presented for accurately transferring salient features from the input medical images into a single fused image. Firstly, the input images are transformed into their multi-scale representations by Laplacian pyramid. Secondly, the contrast feature map and outline feature map are extracted from the images at each scale, respectively. Thirdly, after extracting the multiple features, an efficient fusion scheme is developed to combine the pyramid coefficients. Lastly, the fused image is obtained by a reconstruction process of the inversed pyramid. Visual and statistical analyses show that the quality of fused image can be significantly improved over that of typical image quality assessment metrics in terms of structural similarity, peak-signal-to-noise ratio, standard deviation, and tone mapped image quality index metrics. The contrast is also well preserved by histogram analysis of images.