Medical Image Feature Extraction and Fusion Algorithm Based on K-SVD

In order to better fuse the CT and MR images, based on the classical image fusion method, an image feature extraction and fusion algorithm based on K-SVD is presented. The images are sparse representation. The images are divided into blocks via the sliding window. The dictionary is compiled the column vectors. The redundant dictionary is learned by the K-singular value decomposition (K-SVD) algorithm. Then we solve the sparse coefficient matrix for each original image. And combining sparse coefficient of nonzero elements realizes the image feature fusion. Finally, the reconstructed fusion image is obtained from the combined sparse coefficients and the overcomplete dictionary. The method in this paper is capable of extracting image features and the strong anti noise interference. Experiments show that this method better preserves the useful information in the original image and the fusion image details are clear. Compared with other fusion algorithms, the results show that the proposed method has better fusion performance in both noiseless and noisy situations and is superior.