Image super-resolution reconstruction based on sub-pixel registration and iterative back projection

In respect to a sequence of low-resolution images taken from a moving camera to the same scene, a practical approach is proposed to reconstruct a high-resolution image utilizing the complementary information between these images. The four-parameter rigid transformation model is proposed. The movement parameters are registered by Taylor series expansion. The displacement vectors are estimated from coarseness to fine by the use of gauss pyramid model. According to the estimated sub-pixel precision parameters, image super-resolution reconstruction is performed by the adoption of iterative back-projection (IBP). The experimental results show that the algorithm in this paper achieves high registration precision and good reconstruction effect.