Arbitrary Image Scaling using a Filter Bank with Raised Cosine Function based Sampling Kernels

In this paper, a new approach for arbitrary scaling of images using a 2-D M-channel filter bank is proposed, whereby sampling kernels, introduced along with an optimized adaptive interpolation technique, are incorporated into the filter bank. More specifically, 2-D down and up sampling formulae are derived in a closed form by utilizing an adjustable window function such as a raised cosine function. Also, an optimized adaptive interpolation technique is employed in the synthesis part to compensate for quality degradation arising in scaled images. Simulation results show that high-quality images of arbitrary size can be obtained from the original image by the proposed approach. Key-Words: arbitrary scalability, M-ch DFT filter bank, raised cosine function, optimized adaptive interpolation