Abstract: Two optimized coding methods are presented to reduce the code length of binary image rectangular partitioning compression,where redundant bits appear in representing the posi-tion of rectangular vertices.One approach applies adaptive arithmetic coding with a compressed symbol set,which improves encoding efficiency at a cost of more complex computation.The second is Column Averaged-Adjacent Position Relevancy Encoding(CA-APRE),which reduces code length more than 15%-30% to adequately utilize the transcendental structure information of row and column in a binary image,and keeps encoding and decoding computation the same as before. CA-APRE is apparently better than adaptive arithmetic coding in both computation complexity and encoding efficiency.