New single-image super-resolution reconstruction using MRF model

Super‐resolution image reconstruction using molecular docking

“…Therefore, the training images must have proper generalisation capability [26] and should contain sufficient high‐frequency information w.r.t the input LR image [3]. In [5, 6], the authors proposed mutual information and spatiogram‐based matching criterion to select efficient training image dataset. In [9, 27] authors used norm and diffusion distance of the colour histograms respectively to select efficient training image dataset.…”

“…In this section, effectiveness of the proposed RiBMD‐SR method is validated via several experimental evaluations. Performance analysis of this method is compared with some of the popular state‐of‐the‐art LSI‐SRR methods such as, example‐based super‐resolution (EX‐SR) [1], new super‐resolution method using MRF (NMRF) [6], neighbour‐embedding‐based super‐resolution (NE‐SR) [7], robust neighbour‐embedding‐based super‐resolution (RNESR) [9], sparse‐coding based super‐resolution (SC‐SR) [10], evolutionary sparse‐coding based super‐resolution (ESC) [14], along with two widely used CNN‐based SRR methods [16, 17]. We will name the methods in [16, 17] as CNN1, CNN2, respectively.…”

“…This is due to the fact that this method ignores the geometry structure in the reconstruction process by utilising uniformly generated training patches. As compared to the above‐discussed methods, NMRF method [6] yields enhanced output images. This method is based on the framework of EX‐SR method but utilises spatiogram‐based matching criterion for the selection of potentially efficient training images.…”

“…Existing state‐of‐the‐art LSI‐SRR methodologies fall on the category of Example‐based‐SRR (EXSRR) methods [1–6], Neighbour‐embedding‐based‐SRR (NESRR) methods [7–9] and Sparse‐representation‐based‐SRR (SCSRR) methods [10–14].…”

“…Moreover, convergence rate of this method is very slow. To avoid these limitations and to provide high quality HR solutions, a variety of methods [3, 4, 6] have been proposed in the literature.…”