针对使用传统评价指标依托单一特征来反映修复后古代壁画图像质量的局限性，同时为增加古代壁画修复图像质量评价方法的多元性与丰富性，本文结合敦煌壁画图像颜色特征和纹理特征，提出一种基于多维度特征融合的全参考敦煌壁画修复图像质量评价方法。首先，分别计算原完好壁画图像与修复后壁画图像的颜色熵差、归一化中心距差、颜色直方图的卡方距离、颜色标准差之差值、图像能量差、相关性差、对比度差以及同质性差值；其次，利用主成分分析法利用8个特征属性建模得到质量评价模型，模型的输出值即为修复后图像的质量得分。在有效性验证实验中，该方法的得分获得与其他传统的客观评价方法一致的结果，具有较强的相关性。该评价指标综合分析敦煌壁画图像的色彩分布以及纹理的细节保留和模式的重现，实现了一种全方位、多层次的图像修复质量评价方法。

Aiming at the limitations of using traditional evaluation indexes to reflect the image quality of ancient murals after restoration by relying on a single feature, and at the same time increasing the plurality and richness of the ancient murals restoration image quality evaluation method, in this paper the Dunhuang mural image colour features and texture features were combined, and a full-reference Dunhuang mural restoration image quality evaluation method based on the fusion of multi-dimensional features was put forward. Firstly, the colour entropy difference, normalized centre distance difference, chi-square distance of colour histogram, colour standard deviation difference, image energy difference, correlation difference, contrast difference, and homogeneity difference between the original intact murals image and the restored murals image were calculated respectively; secondly, the eight feature attributes were further modeled into a comprehensive expression to obtain the quality evaluation model by using the principal component analysis, and the output value of the model was the quality of the restored image. In the validation experiments, the scores of this method were consistent with other traditional objective evaluation methods with strong correlation. The evaluation index comprehensively analyses the colour distribution of Dunhuang murals images as well as the detail retention of textures and the reproduction of patterns, realizing an all-round, multi-level image restoration quality evaluation method.