The region based image coding is a new image coding scheme based on the contour-texture modeling.
In this thesis, a new contour coding method is investigated for use in region based image coding. Generally the contour data are represented by chain coding method. But the data compression efficiency is low since contour pels along the border regions are coded twice and heavy burden for initial absolute address representing each contour.
To alleviate the problem, the Depth First Search in Graph traversal algorithm is introduced to the chain coding method. Also, systematic methods are devised to simplify node types, to eliminate needless node (vertex), and to further reduce codeword length.
The proposed coding scheme is shown to be efficient for contour images obtained by split-merge segmentation scheme. For some test image, we can reduce data about 60% with regard to chain code (not coded twice). Also, this coding method is shown to be effective for other binary images such as weather map or geographical map.
