High Accuracy Feature Extraction Using Chain Code in Gray-Value Images
Original Publication Date: 1988-Jun-01
Included in the Prior Art Database: 2005-Feb-15
In order to improve the accuracy of geometric feature extraction of contrasting objects for machine vision, the chain coding scheme, originally designed to encode the boundary line of regions in binary images, is extended to allow for tracing the contour between objects and background in gray-value images with sub-pixel accuracy resulting in increased accuracy in determining geometric object features, such as area and center of gravity. A two-step process is used. First, the chain code of a suitably binarized version of an image is determined. Then, gray-value information from the original image is used to more accurately determine the boundary line and, therefrom, region features. A chain code defines the boundary of an object from a starting point in terms of incremental movement along the boundary.