Image Zooming Method with Hierarchical Structure

Since the enlarged images which were calculated by traditional linear interpolation and nonlinear algorithms are disturbed by blur and fuzzy edge, this paper proposes a new image zooming algorithm with hierarchical structure. It employed the retinex model to decompose image into low frequency layer and high frequency layer. The mirror signals which were introduced by up-sampling course were filtered by low pass filter to reconstruct the low frequency layer. The heat equation was adopted to reconstruct the high frequency layer with edge-preserving ability. The zooming images were obtained by merging layers at last. Experimental results show that this algorithm improves the defects of linear interpolation and nonlinear algorithms, the generated images via our algorithm possess preferable visual effect.