Method of computing sub-pixel euclidean distance maps

摘要

Boundary curves of a source image are identified. Each pixel of a distance map is associated with a corresponding region of the source image. Each pixel is assigned a calculated distance value corresponding to the Euclidean distance between a center of that pixel and the nearest point of the closest boundary curve. The nearest point is located to sub-pixel accuracy. A method of compressing and decompressing a source image includes the step of generating a first distance map having a first resolution. The first distance map is downsampled to generate a second distance map having a second resolution. The second distance map may be interpolated to generate an interpolated distance map having the first resolution. A soft threshold is applied to the interpolated distance map to generate a reconstructed source image having the first resolution.