The computation error is less than one ulp 计算误差小于一个ulp 。
Compared with the traditional algorithm ( computation error is more than 10 percents ) , the computation precision of the proposed algorithm was greatly increased 利用本算法,图像面积和周长的计算误差最大不超过2 。
In numerical computation aspect , by implying new computational geometric field , computation error caused by boundary condition was decreased 在数值计算方面,本文采用了新的计算域,减小了由于边界条件造成的计算误差。