A new contour simplification filter for region-based coding

In region-based (RB) coding, the image is divided into a number of various-shaped regions, which are then treated as objects and coded. Experimental results show that a bottleneck in RB coding at very low bitrates is the amount of shape information, represented by the contours separating the regions, which has to be coded. This paper presents a new filter for contour simplification. The filter reduces the number of contour points by an average of more than 30%. A simplification example shows that its use does not affect subjective image quality.