Image Fusion Method Research Based On Quincunx Sampling Non-separable Wavelets

With the rapid development of sensor, computer technology and information technology, the application of image fusion technology is becoming wider and wider. It can be used in the fields of military affairs, medical science, remote sense, target identification, machine vision. For now, tensor product wavelet transform in the image fusion has been widely used. But it is neither fully using data itself unique features, nor the optimal function representation method. It can catch only limited direction information of horizontal, vertical and diagonal directions, so that it is not fully using image itself geometry regularity. Recently, with the continuous research of image directions and wavelet transform, non-separable wavelet transform and other methods have been proposed, and non-separable wavelet transform can be divided into two-channel, three-channel and four-channel non-separable wavelet transform. They have become outstanding in the study of image directions. This, because of the lowest amount of computation and its excellent performance of the two-channel wavelet decomposition and reconstruction, but also has a wide range of research and application value.In order to solve the problems that the image fusion method based on separable discrete wavelet transform has block effect in the fused image and the spatial resolution is lower in the fusion result image of nonsubsampled non-separable wavelet transform, a new image fusion method based on sampled two-channel non-separable wavelets is proposed. Simultaneously, the filter banks relating to this king of non-separable wavelet are constructed, and the following aspects are made.On the one hand, a multi-spectral image fusion method based on quincunx sampling two-channel non-separable wavelets is presented in this thesis. The multi-resolution decompositions of the intensity of the multispectral image and panchromatic image are performed in quincunx sampled mode using the constructed filter bank. The low frequency and high frequency coefficients are fused by different fusion rules. The experiment results show that this method has good visual effect. The fusion performance outperforms the IHS fusion method, IHS-DWT in preserving spectral quality and high spatial resolution information. When compared with the fusion method based on nonsubsampled two-channel non-separable wavelet, the proposed method has been observed to have higher spatial resolution and good global spectral information.On the other hand, a multi-focus image fusion method based on quincunx sampling two-channel non-separable wavelets is presented in this thesis. The multi-resolution decompositions of different focus images are performed in quincunx sampled mode using the constructed filter bank, then the weighted averaging operator was used to carry on the fusion to the low frequency coefficient, the max of absolute value was used to carry on the fusion to each solution high frequency coefficient. Finally, to reconstruct the coefficient of fusion, the fusion image in which the high and low frequency proportion elevates was obtained. The experiment results show that, the method significantly reduces the RMSE value, improves the PSNR value, and makes that the fused image has good effect.