Content Dependent Quality Measures for Image Fusion

要旨

Satellite images acquired form sensors exhibit either good spectral characteristics (multispectral data) or high spatial characteristics (panchromatic data) in remote sensing. Image fusion is imperative to enhance the interpretation quality of images. The quality evaluation of the decisive fused image is a difficult task in the realm of remote sensing, due to unavailability of a reference image general. An attempt through this study is made, to obtain the objective measurements using content based segmentation for evaluating the performance of the fused images. The spectral and spatial changes for each region and the signal to noise ratio between the original multispectral image and the fused image are used to assess the performance of the fusion technique. Most commonly used fusion techniques such as Principal component analysis, Multiplicative merge, Brovey transform and Lifting Wavelet transform are evaluated using the proposed approach. The results illustrate that the content based segmentation is objective and any priori information is not a requisite. The performance evaluation ascertains that Lifting Wavelet transform outperforms the other fusion techniques.