基于CSSIM的印刷品图像质量评价方法

目的为解决印刷品图像质量评价过程中主观与客观评价结果不一致的问题,以印刷制品为研究对象,基于Contourlet变换提出一种改进的图像质量评价算法。方法在Contourlet域进行多尺度、多方向分解,以获取不同尺度的图像特征。将结构相似度直接应用于各Contourlet分解频带,得到不同频带的结构相似度;对不同频带的结构相似度求加权和,进而获得整幅图像的结构相似度,即图像的最终评价指标。结果以ISO标准测试图像为基础进行实验研究的结果表明,所述方法的综合评价效果最理想,与主观评价的一致性最好。结论基于Contourlet变换域结构相似度的图像质量评价方法能够更好地评价图像质量,符合人眼的视觉特性,可用于印刷品图像处理。

Evaluation Method for Printed Image Quality Based on CSSIM

The work aims to propose an improved image quality evaluation algorithm based on Contourlet transform, in order to solve the inconsistency between subjective and objective evaluation results in the process of evaluating the printed image quality. The image was decomposed in a multi-scale and multi-directional manner in the Contourlet domain to obtain image features of different scales. The structural similarity (SSIM) was directly applied to each Contourlet decomposed sub-band and the SSIM of different frequency bands was obtained. The SSIM of a whole image was achieved on the basis of the weighted sum of SSIM from different frequency bands calculated, namely the final evaluation index of the image. The results of experiment study carried out based on the ISO standard test images showed that, the comprehensive evaluation effect of the described method was ideal and the consistency with the subjective evaluation was the best. The image quality evaluation method based on Contourlet transform domain structure similarity (CSSIM) can better evaluate the image quality and accord with human visual features. Therefore, it can be used for printed image processing.