Image interpolation by using Gaussian regularized regression with cross-based window

To increase resolutions, image interpolation has been widely investigated for several years. Especially, the interpolation techniques for super resolution televisions become more and more important since the most video programs are only with high definition. The linear-based interpolation algorithms bring out the jaggy noise conspicuously. Recently, the new edge-directed interpolation (NEDI) is proposed to improve the accuracy with one-fold training size for predicting parameters. In this paper, an image interpolation based on Gaussian regularized regression with cross-based window (GRR_CW) approach is proposed. The GRR_CW contains spatial confidence consideration and cross-based window generation to lead the prediction more reliable. In experimental results, we prove that the proposed GRR_CW can achieve higher image quality in PSNR and SSIM performances than the traditional linear-based and NEDI-based algorithms.