摘要
针对图像视频处理领域的多参数优化问题和穷举带来的高复杂度,本文提出了一种多参数取值调优方法。受深度学习中的梯度下降思想启发,将参数的复杂度变化和性能变化的比值,即能效比(EER),作为参数候选值优劣的判决依据。基于NLM(non-local means,非局部方法)和迭代的算法是当前图像去噪的一个主流框架。本文提出了一种多参数调优方法,并将该方法应用于该类算法中。实验结果表明,该方法能快速有效地优化算法参数取值,得到目标复杂度约束下的参数最佳取值组合决策,对于算法的实时应用有较大价值。
In view of the multi-parameter optimization problem in image/video processing and the high complexity brought by exhaustion,a novel multiple-parameter optimization method is proposed.Inspired by the idea of gradient descent in deep learning,energy efficiency ratio(EER)which is the ratio of quality and complexity,is used as the basis decision for the parameter candidate.NLM(non-local means)and iterative algorithms are a mainstream framework for current image de-noising.This paper proposed a multi-parameter optimization method and applied it to this kind of de-noising algorithm.The experimental results show that the proposed method can optimize the parameters effectively and obtain the optimal parameter combinations under the target complexity constraint,which has great value for the real-time application of the algorithm.
作者
范梦婷
潘晨(指导)
殷海兵(指导)
FAN Mengting;PAN Chen(Communication Fundamental and System Lab,China Jiliang University,Hangzhou 310018,China)
出处
《电视技术》
2018年第11期11-15,39,共6页
Video Engineering
基金
有记忆信源率失真模型及高并行度HEVC编码算法优化方法(61572449)
关键词
多参数优化
能效比
复杂度
图像去噪
multiple-parameter optimization
energy efficiency ratio
complexity
image de-noising