Thyroid Image Fusion Based on Shearlet Transform and Sparse Representation
ZHENG Wei;SUN Xueqing;HAO Dongmei;WU Songhong;College of Electronic and Information Engineering,Hebei University;Key Laboratory of Hebei on Digital Medical Engineering;Affiliated Hospital,Hebei University;
According to the characteristics of ultrasound images with low contrast and SPECT images with blurred boundary, combining the theory of multi-scale geometric analysis with single scale sparse representation, an image fusion algorithm based on Shearlet transform and sparse representation is proposed. Firstly, the Shearlet transform is used to decompose the registered source images, thus the low frequency sub-band coefficients and high frequency sub-band coefficients can be obtained. The low frequency sub-band coefficients with lower sparseness are used to train the dictionary and the sparse representation coefficients are calculated by the trained dictionary, and the fusion rule of the sparse representation coefficients is used to select the larger energy. The high frequency sub-band coefficients are fused by the region sum modified laplacian. Finally, the fused image is reconstructed by inverse Shearlet transform. The experimental results demonstrate that the proposed method outperforms the multi-scale methods and the methods of sparse representation in single scale in term of visual quality and objective evaluation.