A HYBRID IMAGE FUSION APPROACH BASED ON DCT AND PCA

Abstract

This thesis work, presents A Hybrid Image Fusion Approach based on DCT and PCA. In this, a multisensory image fusion scheme based on the Directional DCT combined with PCA has been implemented and assessed. The source image were divided in to non overlap square blocks of equal size and fusion was carried out in the respective blcoks. This approach works in two levels where in the first level, mode 0 - 8 were executed on the source images. The coefficient from the input images are utilised in the fusion procedure, for every mode. Similarly, procedure is continued for remaining mode as well. The fusion procedure is carried out using the 3 diverse fusion rules i.e., 1. Averaging of respective coefficient (DDCTav), 2. Selecting respective frequency bands with the maximum energy (DDCTek) and the 3. Selecting respective coefficients having maximum absolute values (DDCTmx) among the images. Later, in the next level, we get 8 different fused images which are generated one from every mode. The second level begins with fusion of these eight generated images by PCA. The fused output image obtained from PCA is further enhanced by passing through an averaging filter which gives the final fused output image. Performance of the algorithm were compared using fusion quality evaluation metrics such as RMSE, PSNR, S.D., Entropy, etc. It was concluded that though the overall performance of the algorithm is mediocre when compared to other traditional fusion schemes, but enhanced output showed improvements in the implementation resuts in terms of the performance metrics.