2D Image Reconstruction and Zooming with Simultaneous Edge Preservation and Blur

This paper presents a method of reconstructing twodimensional images at varying spatial resolutions while preserving edge information and allowing for blurring. It is often the case that a user would like to zoom in on certain areas of an image. Such a zoom operation requires a reconstruction of the data to provide values at the resampled pixel locations. Standard reconstruction methods use a spatially invariant reconstruction kernel such as a Gaussian or spline to perform this resampling. These reconstructions result in blurred edges and preserve noise artifacts. The reconstruction method described in this paper uses a spatially varying kernel to allow preservation of edges at the various resolutions. In addition the reconstruction allows the user to specify regions to blur in the reconstructed image. This allows the user to display magnified images in which the edge locations are preserved and unimportant areas are deemphasized. This reconstruction scheme has been applied to several test data sets. The resulting images demonstrate the algorithm’s ability to magnify an image while preserving edges and blurring unwanted detail.