故事背景
本文方法（DBIQ）

发表在2016年Neurocomputing。

摘要

JPEG is the most commonly used image compression standard. In practice, JPEG images are easily subject to blocking artifacts at low bit rates. To reduce the blocking artifacts, many deblocking algorithms have been proposed. However, they also introduce certain degree of blur, so the deblocked images contain multiple distortions. Unfortunately, the current quality metrics are not designed for multiply distorted images, so they are limited in evaluating the quality of deblocked images. To solve the problem, this paper presents a no-reference (NR) quality metric for deblocked images. A DeBlocked Image Database (DBID) is first built with subjective Mean Opinion Score (MOS) as ground truth. Then a NR DeBlocked Image Quality (DBIQ) metric is proposed by simultaneously evaluating blocking artifacts in smooth regions and blur in textured regions. Experimental results conducted on the DBID database demonstrate that the proposed metric is effective in evaluating the quality of deblocked images, and it significantly outperforms the existing metrics. As an application, the proposed metric is further used for automatic parameter selection in image deblocking algorithms.

结论

Image deblocking has been extensively researched for removing blocking artifacts in JPEG images. However, the quality evaluation of such deblocked images is still an open problem. In this paper, we have presented a no-reference quality model for evaluating the quality of deblocked images. Blocking artifacts in smooth regions and blur effects in textured regions are considered in the proposed model. It is a moment-based metric, where the Tchebichef moments are used to achieve: (1) block classification, (2) blocking artifact evaluation, and (3) blur evaluation. We have also built a deblocked image database DBID to compare the performances of image deblocking algorithms, and also to verify the performance of the proposed method. The experimental results have demonstrated that the proposed method is effective in evaluating the quality of deblocked images, and it significantly outperforms the state-of-the-art blocking artifact metrics, blur metrics and general-purpose NR image quality metrics. As an application of the proposed model, we have also used DBIQ for automatic parameter tuning in image deblocking algorithm, producing very promising results.
In this work, the proposed quality model is based on the Tchebichef moments of gray-scale images. However, color also affects the quality of deblocked images, so the performance of the proposed metric could be further enhanced by incorporating color information. A straightforward way to improve DBIQ is to use quaternion-type moments [51]. Furthermore, the presented work mainly focus on deblocked JPEG images. More general deblocking scenarios, e.g., deblocking loop filter in H.264/AVC, will be investigated in future work.

要点

我们讨论JPEG压缩图像的块效应。
许多去块效应方法都会引入模糊，导致图像中存在多重失真。然而，现存质量指标都只局限于单一失真（比如块效应），没有考虑模糊等其他失真。
为了解决这一问题，本文针对去块效应的图像（deblocked images），提出了一种无参考质量评价指标：NR DeBlocked Image Quality（DBIQ）。
方法核心：同时评估平滑区域的块效应，以及纹理区域的模糊程度。
具体而言，作者采用了切比雪夫矩（Tchebichef moments），同时实现了块分类、块效应评估和模糊程度评估。
本文还有建库等贡献。作者将预测的指标用于去块效应算法，发现实验结果有所提升，证明了该指标的有效性。
局限：只考虑了灰度图像；只考虑了JPEG图像。

故事背景

作者观察了借助去块效应算法[14]得到的图像，发现：去块效应图像的平滑区域容易遭受（残留）块效应影响，而纹理区域容易变模糊。

The deblocked images are contaminated by both blocking artifacts and blur. Blocking artifacts mainly affect the quality of smooth regions and blur mainly affects the quality of textured regions.

基于此观察，作者提出用离散切比雪夫矩，同时评估块效应和模糊。

本文方法（DBIQ）

DBIQ由两部分组成：

块效应指标RMB[26]，是当时最好的检测块效应指标；
基于矩的模糊检测模块。

整体框图：

流程大致如下：

首先，deblocked图像被不重叠地分为\(8 \times 8\)的目标块。\(8 \times 8\)应该是JPEG编码块的尺寸。
对于每一个目标块，计算切比雪夫矩。根据非直流分量的平方和（the sum of squared non-DC moment, SSM），决定该块的类型：平滑还是纹理。
对于平滑块，我们通过RMB方法计算块效应指数。
对于纹理块，我们通过本文的方法计算模糊效应指数。作者还加入了显著性图。
两个得分通过池化，得到最终得分。

博主更关注块效应的检测，因此跑去看块效应质量评估论文啦。

巴特西

Paper | No-reference Quality Assessment of Deblocked Images

故事背景

本文方法（DBIQ）

最新文章

热门文章