Region Based Image Fusion Using Modified Contourlet Transform

Image fusion techniques are applied in various fields such as remote sensing, medical imaging, concealed weapon detection, etc. Combining two or more images of the same scene usually produces an output image which provides increased interpretation capabilities and reliable results. In image fusion, data with different specifications such as resolution, spectral and spatial coordinates are combined. Image fusion algorithm can be categorized into pixel and feature levels. Region based method is one way of achieving the feature- level fusion. Segmentation plays a vital role in this fusion process where the features of the source images are extracted first using Edge based segmentation Consequently, the Contourlet transform is applied on the different regions and the coefficients from different regions are merged separately. Finally, the fused image is obtained by performing inverse Contourlet transform. The Laplacian pyramid employed in Contourlet transform is not the perfect transform from the point of view of image fusion, since it involves down-sampling procedure which makes it shift variant. Therefore, in order to yield better performance metric in the proposed work, the Contourlet transform is modified by replacing the Laplacian pyramid by Contrast pyramid. Region based image fusion using modified Contourlet transform and the Contourlet transform are applied on various images to compare their performances. Simulation results indicate that Region Based Image Fusion using Modified Contourlet transform produces better results than Contourlet transform in terms of entropy, correlation coefficient, PSNR and average gradient.