Optimal bit allocation for maximum absolute error distortion in the application of JPEG2000 part 2

This paper proposes a strategy to deal with bit rate allocation for the optimization of maximum absolute error (MAE), or 1-infinity, distortion metric in 3D data compression using JPEG2000. Part 2 of this standard has the capability to compress 3D data by treating data as separate 2D slices; these slices could be taken directly from the data or from the data after it has undergone a decorrelation transform (KLT) in one direction. To perform bit rate allocation we use a mixed model approximation to the MAE rate-distortion curve, that is used in an optimization algorithm. To solve the problem of MAE-related bit rate allocation in the KLT domain, we theoretically derive an upper bound for MAE based on the basis vectors of the KLT; we also develop an algorithm for optimizing this upper bound, and we illustrate how the minimization of this upper bound can decrease the actual MAE.