Abstract:

As the existing fusion algorithms cannot effectively overcome the contradiction between the strong denoising ability and the detailed information reservation, a new fusion algorithm based on the characteristics of wavelet transform modulus maximum is proposed. In the proposed algorithm obtained from the adaptive weighted averaging of the window-region intensity, first, the characteristics of wavelet transform modulus maximum in different scales are extracted with a chosen wavelet radix. Next, the noise is filtered according to the different characteristics of Lipschitz index between the signal and the noise at local singularity points. Then, the local region intensity of the modulus maximum is calculated, and the weights of wavelet coefficients are dynamically distributed in the sub-images in different scales. Finally, the sub-images are reconstructed to obtain a fused image. The fusion experiments of computed tomography and positron emission tomography images indicate that the proposed method can adapt itself to various fusion demands and can restrain the noise with the detailed information being reserved.

Key words: wavelet transform, image fusion, adaptive algorithm, modulus maximum, local region intensity