An improved image superresolution using local adaptive unsharp masking

The proposed work presents an efficient, local adaptive sharpening scheme which sharpens the low resolution images so as to effectively reduce the blurring in the up-sampled, high resolution images. The up-sampled, high resolution images or video intra-frames generated using various interpolation techniques, suffer through blurring at the edges and fast changing regions. The degree of blurring is more in the high variance regions as compared to the slowly varying, flat regions. Therefore, to restore the high frequency information effectively, the high variance regions should be more enhanced than the flat or slowly varying regions prior to the up-sampling process. So, in this proposed method, the blurring kernel, used in unsharp masking operation, is made locally adaptive as per the normalized statistical local variance in a 3×3 neighborhood. As per the adaptive algorithm, the central weight of the blurring kernel varies directly as per the local descriptor of a 3×3 neighborhood. Consequently, the high variance regions are more enhanced than the slowly varying flat regions to counter the blurring problem in the up-sampled, high resolution images or video sequences. Experimental results show that the proposed method gives much better performance than the widely used interpolation schemes in terms of objective and subjective measures under varying zooming conditions.