融合美感注意力的图像美学评估算法

An image aesthetic evaluation algorithm by integrating aesthetic attention

图像美学评估是图像处理领域的研究热点。现有方法不能较好的描述人类视觉系统对图像美学的判断过程,因此研究了融合美感注意力的评估算法。首先,采用弱监督学习的方法提取美感注意力特征图,该方法能够克服现有方法未充分利用人类视觉系统特性以及现有数据集中缺少美感注意力标签的问题。然后,将美感注意力特征图与深度特征图进行交叉融合,共同构建美学分数预测模型。实验证明,该算法相比现有算法具有更优越的性能表现。

Image aesthetic evaluation is a research hotspot in the image processing field, but the existing methods cannot better describe the characteristics of human visual system. In this paper, an evaluation algorithm integrating aesthetic attention is proposed.First, a weakly supervised learning method is used to extract the aesthetic attention map, which can overcome the problems that the existing methods do not fully utilize the characteristics of the human visual system and the existing datasets lack aesthetic attention labels. Then, the aesthetical attention map is fused with the deep feature map to construct the aesthetical score prediction model. Experiments demonstrate the superiority of the proposed method to the existing approaches.