An enlargement method of digital images based on Laplacian pyramid representation without edge-effect

摘要

It is necessary to predict unknown higher-frequency components which are lost by sampling for enlarging digital images. Based on Laplacian pyramid representation, the prediction of unknown higher-frequency components is equivalent to the prediction of an unknown high-resolution Laplacian image. Greenspan et. al., proposed the image enlarging method based on this Laplacian pyramid representation. This method has two free parameters and Takahashi et. al., found suitable these two parameters by seven natural digital images. However, the artifacts are appeared near the step-edge regions of the enlarging result with two parameters. In this paper, we derive suitable two parameters for the step edge signal. We show the step edge type signal is detected by using the two layers of Laplacian images. Thus, we can detect the step edge signal regions into the processed image and we apply the two parameters for the step-edge signal at these regions. Excellent enlarging images without artifacts near the edges are obtained by the novel method.