An Efficient Algorithm for Image Scaling with High Boost Filtering

Image scaling is the process of resizing a digital image, wherein an image is converted from one resolution/dimension to another resolution/dimension without losing the visual content. It has many terminologies in literature such as Image Interpolation, image re-sampling, digital zooming, image magnification or enhancement, etc [1] [2]. Image interpolation algorithms can be grouped in two categories, non-adaptive and adaptive [3] [4]. Interpolating from lower to higher resolution it is termed as up-scaling / up-sampling and from higher to lower resolution it is termed as down-scaling / down-sampling. In this article we propose a non-adaptive image interpolation algorithm to scale images for any given scaling ratio with an enhancement scheme to ensure a better image quality metric (PSNR in dB) of the scaled image. In the proposed algorithm, two reference images, one with higher resolution and another with lower resolutions are generated using the original input image and scaling factor. Later, two Intermediate images are generated from these reference images, using the filtering based re-sampling Lanczos3 kernel. Finally, the desired scaled image is obtained by linearly interpolating the two intermediate images. The desired scaled image is then passed through an enhancement phase, where a High Boost Filter is employed, to obtain an enhanced scaled image. Finally the algorithm is compared with the other interpolating techniques such as Bilinear Interpolation, B-Spline interpolation, Lanczos interpolation etc. in terms of the Image Quality metric: PSNR in dB.