Error based Sharpening for Efficient 2-D Up-scaling Performance

This paper presents a spatial domain, prediction error based global sharpening scheme to effectively handle the blurring problem in a 2-D up-scaled image or video intra-frame. Most of the interpolation schemes available in the literature produce blurring artifacts in an up-scaled image because of their analogy with an LPF operation. The degradation of very high and ultra high frequency components is much more than the medium and high frequency components. So, enhancement of very high and ultra high frequency components prior to image up-scaling would give much less blurring in the up-scaled images. The proposed prediction error based pre-processing scheme is used to enhance the high and very high frequency regions of a low resolution image or video intra-frame so as to effectively reduce the blurring in its corresponding up-sampled, high resolution counterpart. The proposed scheme extracts the degraded high frequency information by purposely down sampling the given low resolution image at a very low compression ratio and then restores it back to its original size using Bicubic interpolation. So, the prediction error significantly contains the lost very high and ultra high frequency information. So, the weighted superimposition of the prediction error with the original image prior the up-sampling process reduces the blurring considerably in the up-sampled image. Experimental results show the proposed method gives better objective and sUbjective performance than the other widely used interpolation schemes.