Image resolution enhancement using inter-subband correlation in wavelet domain

Abstract

Image resolution enhancement in wavelet domain has been researched for years and proved successful. Compared with the conventional bilinear and bicubic interpolation method, the wavelet-based resolution enhancement methods enhance the resolution by estimating the preserved high frequency information in the given images after inverse wavelet transform. Although several wavelet-based methods have been proposed for resolution enhancement, a new inter-subband based method is proposed, which shows better performance than the previous methods. Our proposed inter-subband image resolution enhancement algorithm utilizes the correlation between subbands having different sampling phase shifts. By analyzing correlations between subbands in the low level, transition filters between subbands are designed and applied to the subbands in the high level. The filter is estimated under the assumption that filters connecting two subbands are very similar in different levels. The experimental results show that our proposed method outperforms the conventional interpolation methods and competing wavelet-based method with respect to peak signal-to-noise-ratio(PSNR) as well as the subjective quality. Moreover, our assumption that filters connecting two subbands are almost same in different levels has been confirmed in our experiments.