Infrared Image Super-resolution Reconstruction Based On Sparse Representation

Infrared imaging is a technique that can convert an infrared radiation from an object itself into an image by using an infrared imaging device. Therefore, the infrared imaging device can work in the night and harsh environment, and has widespread use in many fields. However, the resolution of the infrared image is usually lower as limited by the resolution of the infrared imaging system. Considering the constraints from the hardware implementation and cost, can use super-resolution reconstruction algorithm to improve the resolution of the infrared image.Aiming at the shortcomings of the existing super-resolution reconstruction algorithms, this paper proposed two algorithms for infrared image super-resolution reconstruction based on sparse representation. Specific work is as follows:(1)Aiming at the low efficiency of traditional image super-resolution reconstruction algorithm based on sparse representation, proposed a fast infrared image super-resolution reconstruction algorithm based on sparse representation. Firstly, considering the characteristics of the infrared image and the efficiency of the algorithm, used the LOG operator to extract the feature of the low resolution infrared image. Secondly, used the K-SVD algorithm to update the atom of the dictionary and used the gOMP algorithm to sparse decompose. Finally, used the IBP algorithm to further optimize the reconstructed high-resolution infrared image. The experimental results showed that the quality of the reconstructed image obtained by the proposed algorithm is similar to the traditional super-resolution reconstruction algorithm, but the efficiency of the proposed algorithm is 11 times higher than that of the traditional algorithm.(2)Aiming at the low efficiency of traditional image super-resolution reconstruction algorithm based on sparse representation and the poor quality of the reconstructed image by the algorithm in (1), proposed an infrared image super-resolution reconstruction algorithm based on cross-correlation clustering. Firstly, combining the efficiency of the dictionary training with the validity of the dictionary, proposed an adaptive clustering algorithm based on correlation coefficient. Used the proposed adaptive clustering algorithm to divide the dictionary training sample set into 15 categories. Secondly, for each sub-category training sample set, used the Lagrange dual method to update the atom of the dictionary, and used the FsS algorithm to sparse decompose. Finally, the low-resolution input sample selected the sub-dictionary adaptively in the reconstruction according to the principle, which selected the sub-dictionary corresponding to the maximum correlation coefficient between the low-resolution clustering centers of the sub-category. The experimental results showed that the proposed algorithm not only ensure the better quality of the reconstructed image, but also the efficiency of the proposed algorithm is 6 times higher than that of the traditional algorithm.Both of these two proposed algorithms are more real-time and can improve the resolution of the infrared image effectively. The two algorithms can be used for the image clarification processing in the infrared thermal imaging and have a wide application prospect.