【作者】 刘萍；

【导师】 王啟军；

【作者基本信息】 安徽大学 ， 工程（专业学位）， 2022， 硕士

【摘要】 由于科学技术的迅速发展,智能手机、数码相机等电子设备在人们日常生活中获得了广泛应用,从而形成的图像数据量也大幅增长,而这种海量的图像数据也为磁盘储存和互联网数据传输提出了巨大的挑战。在这种背景下,人们要求图像压缩编码技术不仅要保持图像的质量,而且要逐步降低图像储存的数据量,因此对图像压缩编码技术进行优化具有重要的意义和价值。JPEG作为静态图像最普遍的压缩标准,已被广泛用于各种应用程序和设备,在图像应用领域仍然处于主导地位。对于JPEG编码器,在保证其兼容性的前提下进行编码效率的优化,具有广泛的应用前景。在不确定环境下,实际应用对码率的需求并不明确,需要从同一个原始图像生成多个不同码率的压缩图像以适应多种码率需求,大部分JPEG量化表优化算法中并没有考虑这个问题,而是通过多次优化适配多种码率,效率非常低。从这个角度考虑,将多目标优化与JPEG量化表优化相结合,实现了一次优化产生多个最优的量化表,是应对不确定环境下多种码率需求的有效方式。因此,本文提出了基于多目标进化算法框架求解JPEG图像压缩量化表优化方法。本文的主要研究工作如下:(1)本文提出了一种基于率失真最优进化算法求解多码率情况下JPEG量化表优化方法(RDOEA),算法的主要思想是将多目标优化与JPEG标准相结合。与现有的优化方法不同的是,我们的方法充分考虑了率失真优化原理,并提出了几种策略来解决应用中的多码率需求问题。在多目标优化框架中,通过查找表可以有效地评估每个量化表的适应度值,该查找表是基于每个DCT频带的统计信息以预先定义的方式构建的。然后,提出了基于率失真最优原理的种群更新策略,以指导其向最佳码率失真性能的方向进化。此外,为了保持种群的多样性和均匀性,提出了一种基于凸包的环境选择和局部搜索策略,该策略保留位于第一帕累托前沿面凸包上的解,并通过局部搜索策略进一步丰富空白码率范围的解。实验结果证明了我们的方法在几个经典数据集上解的分布性、编码效率和计算复杂度方面更具有优越性。(2)本文提出了一种基于码率约束和视觉主观质量的JPEG图像编码量化表优化算法。在RDOEA方法中优化目标均采用均方误差(Mean Square Error,MSE),该指标与图像主观质量并不一致,而且JPEG标准通过质量因子调整不同的压缩率,但是不同质量因子控制得到的码率分布不均匀,针对这些问题,提出了基于码率约束和视觉质量的JPEG编码进化优化算法,采用双种群优化框架,用简单问题辅助复杂问题,提高复杂问题的收敛速度。优化过程中将DCT子代相似度(DCT Subbands Similarity,DSS)作为图像视觉质量度量指标,平衡约束与目标适应度值,提升JPEG编码效率。在Kodak数据集上的实验结果验证了算法的有效性。更多还原

【Abstract】 Due to the rapid development of science and technology,electronic devices such as smart phones and digital cameras have been widely used in people’s daily life,resulting in a substantial increase in the amount of image data,and this massive amount of image data poses a huge challenge to data storage and transmission.Under this background,it is required that the image compression technology can further reduce the data amount while image quality can be retained.Therefore,it is of great significance and value to optimize the image compression technology.As the most common compression standard for static images,JPEG has been used extensively across applications and devices to compress images,and it is still dominant in the field of image applications.For the JPEG encoder,optimizing the encoding efficiency under the premise of satisfying its compatibility has a wide range of application prospects.In the uncertain environment,the bit rate requirement is not clear,and it is necessary to generate multiple compressed images with different bit rates from the same original image to meet the requirements of various bit rates.However,most of the JPEG quantization table optimization algorithms do not consider this problem,and they are executed several times to adapt to a variety of code rates,and hence the optimization efficiency is very low.From this point of view,the integration of multi-objective optimization with JPEG quantization table optimization can achieve several optimal quantization tables in a optimization run,and these optimal quantization tables correspond to different bit rates which can effectively cope with the ambigous rate requirements in the uncertain environment.Therefore,in this thesis,the research of multiobjective evolutionary algorithm framework is conducted to obtain the optimal quantization tables of JPEG image compression.The main research work of this paper is as follows:(1)In this thesis,rate–distortion optimal evolutionary algorithm(RDOEA)for JPEG quantization with multiple rates is proposed.It is based on multi-objective optimization and is JPEG standard-compatible.Unlike the existing optimization methods,our method fully considers the rate–distortion optimal principle and provides several optimal solutions to address the multiplerate requirement in applications.In the multi-objective evolutionary optimization framework,the fitness of each quantization table for JPEG compression is evaluated efficiently by the searching in a look-up table,which is constructed based on the statistics of each DCT band in a pre-defined manner.Then,the population update strategy based on the rate–distortion optimal principle is recommended to guide the evolution toward the best rate–distortion performance.Furthermore,to maintain the population’s diversity and uniformity,convex-hull based environmental selection is recommended to identify the solutions at the first Pareto front,and the identified solutions are enriched further with scaling operation to fill the blank rate range.The experimental results for several classic datasets demonstrate the superiority of our method in terms of solution distribution,coding efficiency,and computational complexity.(2)In this thesis,the quantization table optimization algorithm for JPEG image coding based on bit rate constraints and visual subjective quality is proposed.In the RDOEA method,the optimization target is Mean Square Error,which is not consistent with the subjective quality of the image,and the JPEG standard adjusts different compression rates through quality factors,but the bit rate distribution obtained by different quality factors is uneven.The evolutionary optimization algorithm of JPEG encoding based on bit rate constraints and visual quality adopts a dual-population optimization framework,assists complex problems with simple problems,and improves the convergence speed of complex problems.In the optimization process,DCT Subbands Similarity is used as an image visual quality metric to balance constraints and target fitness values to improve JPEG encoding efficiency.The experimental results verify the effectiveness of the algorithm on the Kodak dataset.更多还原