Producing a rendered image version of an original image using an image

Abstract

The invention automatically produces a rendered image version of an original image that accurately represents perceptually distinguishable objects, surfaces and edges in the original image. The original image is first converted to a unique data structure representation called an image structure map (ISM) that accurately, compactly and explicitly represents the geometry, topology and signal properties of perceptually distinguishable regions in the original image. The ISM is a dynamic partition of the image induced by a set of image region boundaries, and includes, for each region in the original image, a region data item that indicates the region boundary data items that form the region and a signal property descriptor indicating the value of signals computed for image locations in the region. An automatic segmentation operation determines image region boundaries occurring in the original image by analyzing discontinuities in signal values of original image locations. The topology of the rendered image version is accurate because vertices (endpoints and intersections) of image region boundaries are represented in the ISM in machine representable values that result from modifying exact vertex values using a rounding operation that mathematically ensures topological and geometric consistency with the original image. The method is particularly useful for producing rendered image versions of full color, continuous tone photographic images, and for producing line drawings of such images. The ISM representation of an image is, in most cases, more compact than its original raster form, and serves as a useful representation for archiving the image.