论文部分内容阅读
全连接的玻尔兹曼机模型可全面描述稀疏系数间统计依赖关系,但时间复杂度较高.为了提高基于玻尔兹曼机的贝叶斯匹配追踪算法(BM-BMP)的重构速度和质量,本文提出一种改进算法.第一,将BM-BMP算法的最大后验概率(MAP)估计评估值分解为上一次迭代的评估值与增量,使得每次迭代仅需计算增量,极大缩短了计算耗时.第二,利用显著最大后验概率估计值平均的方式,有效近似最小均方误差(MMSE)估计,获得了更小的重构误差.实验结果表明,本文算法比BM-BMP算法的运行时间平均缩短了73.66%,峰值信噪比(PSNR)值平均提高了0.57 d B.
The fully connected Boltzmann model fully describes the statistical dependencies among sparse coefficients, but the time complexity is high.In order to improve the reconstruction speed of the Boltzmann-based Bayesian Matching Pursuit (BM-BMP) algorithm, And quality, an improved algorithm is proposed in this paper.Firstly, the maximum a posteriori (MAP) estimation evaluation value of BM-BMP algorithm is decomposed into the evaluation values and increments of the previous iteration so that only the increment , Which shortens the computation time greatly.Secondly, the MMSE estimation is effectively approximated by using the average of significant maximum a posteriori (MAP) estimates, and the smaller reconstruction error is obtained.The experimental results show that the proposed algorithm Compared with the BM-BMP algorithm, the running time of the algorithm is shortened by 73.66% on average, and the peak PSNR value is increased by 0.57 d on average.