A Comparative Study on Wavelet based Pixel Level Image Fusion

J. Tian, Y. Zheng, and L. Shen (PRC)


Image fusion, wavelet transform, quality indices


This paper presents a comparative study on wavelet based pixel level image fusion. First a framework for wavelet based image fusion is described and different alternatives for each block in this framework are introduced. We mainly study four types of wavelet transforms applied to image fusion, i.e., DWT (Discrete Wavelet Transform), SIDWT(Shift Invariance Discrete Wavelet Transform), DT CWT(Double Tree Complex Wavelet Transform) and LSQG (Lifting Schemes on Quincunx Grids). Then several quality indices are used to measure the performance of different fusion schemes. These indices do not require any ground-truth or reference images. The experiments results show that fusion schemes based on SIDWT provides a better performance than those based on other wavelet transforms. But the higher performance is at the cost of higher redundancy caused by discarding of the sub sampling. Fusion methods based on DT-CWT and LSQG are good alternatives.

Important Links:

Go Back