A New Residual Image Sharpening Scheme for Image Up-Sampling

The main challenges in image interpolation approaches are edge restoration and preserving texture details while producing high resolution (HR) images. Interpolation causes approximation error in terms of blurring because it predicts unknown pixels in any location of the image grid. To address the aforementioned issue, the suggested method employs residual image sharpening based on the high frequency (HF) details of a low resolution (LR) image. The sub-sample image edge is determined before upscaling with lanczos interpolation. To determine the non-uniform blurring effect in an upscaled image, the HR image is sub-sampled once more, resulting in an LR image denoted as (LR2). The difference between the LR and LR2 images is used to identify the degradation in HF due to upscaling. To compensate for interpolation loss, the edge details of the subsampled image is aggregated to the degraded image to form the edge-residual image, which is then up-sampled. To extract fine details from the interpolated residual image, a high order filter is used. The weighed sharpened image is aggregated with the previously obtained up-sample image to produce an image with preserved edge and texture. The intended algorithm’s performance is measured using images from various databases and compared to state-of-the-art techniques. The proposed scheme outperforms the existing interpolation schemes.