An adaptive split-and-merge method for smoothing and compression of image contours

Abstract

The contour of a digital image usually has a large number of edges and may suffer from quantization error and noise. In feature extraction or shape matching algorithms, contour smoothing must be performed to reduce noise and quantization error and to compress the data while still keeping its original shape. This study presents a split-and-merge method with adaptive tolerance value for smoothing image contours. The tolerance value depends on the grid constant D and the length of line L in collinearity tests. Experimental results on real binary contours show the method is very effective and precise for smoothing of binary image.