Image compression utilizing the DTOCS, topological raster patterns and delaunay triangulation

In this paper, a new image compresion method is presented. It is based on the use of the Distance Transform on Curved Space (DTOCS), which calculates an integer approximation of the weighted pseudo-Euclidean distance transform along discrete 8-paths, and topological raster patterns to determine coordinates for control points. The decompression is based on triangulating among the control points using Delaunay triangulation. The obtained images are compared to previously published methods and to the JPEG DCT-based method. It is shown that the obtained results are of the same quality as JPEG results and exceed other DTOCS and Delaunay based methods. It is also shown experimentally that the time complexity of the compression method presented in this paper does not depend on the number of control points, i.e. the compression ratio. Because of this and the locality of the decompression scheme this method could well be used in applications where a rapid flashing image is made more accurate over time.