融合人眼掩蔽效应和图像梯度的块效应评价方法

A Blocking Artifacts Evaluation Method Integrating Human Eye Masking Effect and Image Gradient

针对JPEG格式的图像采用分块离散余弦变换的压缩方式易产生块效应的情况,提出了一种高效的无参考块效应评价方法.首先,对图像块边界处像素点的梯度进行变换得到图像块效应映射图(主要包括图像块效应边界的位置和强度信息);然后,计算人眼对图像的亮度和纹理掩蔽效应,将其结合到块效应映射图中,得到显著性块效应映射图,并使用Minkowski法计算出图像块效应评价指标;最后,在LIVE等图像质量评价数据库中进行了大量的试验仿真.仿真结果中单调一致性指标——SROCC和KROCC分别达到了0.9、0.7以上.

Blocking artifacts were the result of block-based discrete cosine transform in JPEG coding.An efficient no-reference method to measure the blocking artifacts was proposed in this paper.Firstly,the gradient of the pixels at block boundaries was transformed into the blocking artifact map,which mainly included the positions and intensity information of blocking artifact boundaries.Then the effects of luminance and texture masking on blocking were figwredout and integrated into the blocking artifact map to form an noticeable blocking artifacts map.Based on the noticeable blocking artifacts map,the Minkowski method was used to calculate the metric of image blocking artifacts.Finally,a large number of experimental simulations were performed in LIVE and other image quality assessment databases.The simulation results showed that SROCC and KROCC reached 0.9,0.7 or more.