Text images superresolution and enhancement

Stroke width‐based directional total variation regularisation for document image super resolution

“…Results of 2× SR using different regularisation terms, in different SR methods. The input LR image is from the dataset I, Times New Roman, 10 pt, and at 75 dpi resolution a Bicubic interpolation, PSNR = 16.18, SSIM = 0.781 b Huber and bimodality [2], PSNR = 16.55, SSIM = 0.840 c BTV and bimodality [15], PSNR = 16.90, SSIM = 0.843 d Proposed SWDTV, PSNR = 17.48, SSIM = 0.862…”

“…Results of 4× SR using different regularisation terms, in different SR methods. The input LR image is a portion of LR Persian handwritten text taken by a cell phone camera a Tikhonov and bimodality [11] b Huber and bimodality [2] c BTV and bimodality [15] d Proposed SWDTV…”

“…PSNR and SSIM curves of different methods in the presence of noise. Methods and their regularisation term(s) are bicubic interpolation, Huber and bimodality [2], BTV [3], BTV and bimodality [15], DBTV [14], and the proposed SWDTV a PSNR values for salt and pepper noise with different densities b SSIM values for salt and pepper noise with different densities c PSNR values for zero mean Gaussian noise with different variances d SSIM values for zero mean Gaussian noise with different variances…”

“…Our proposed regularisation term adaptively performs smoothing based on the local width and local direction of the character strokes, and it is called the stroke width‐based directional total variation (SWDTV). The SWDTV regularisation term can be utilised in single [4–12] or multiple [2, 3, 13–15] document‐image SR methods, as well as in example‐based [4, 5, 7, 9, 10, 12] or non‐example‐based [2, 3, 6, 8, 11, 13–15] document‐image SR methods.…”

Deep Laplacian Pyramid Network for Text Images Super-Resolution

“…Text image super resolution. Multiple-frame SR for text images are addressed in [14,15]. Capel et al [14] utilize a MAP approach with Huber prior, which encourages a piecewise constant solution.…”