【摘要】 绘画风格的多样性为构建艺术形象提供了丰富的视觉信息。为解决风格迁移的色彩单一性和色彩溢出问题,提出一种基于深度学习的图像多风格融合算法。该算法利用预训练的VGG19Net提取网络各层特征,将内容图与多个风格图进行分离重组,构建了新的风格损失函数,并将多个风格特点融合,得到新的艺术风格图片。实验结果表明,多风格融合的新图片中包含多种单风格信息,使转换后的新图片拥有更丰富的视觉信息。基于深度学习的图像多风格融合算法构建了更加丰富的视觉信息,为新的艺术创作提供了参考。更多还原

【Abstract】 The diversity of painting styles provides rich visual information for the construction of artistic images. In order to solve the problem of color monochromatism and color overflow in style transfer,this paper proposes an image multi-style fusion algorithm based on deep learning. This algorithm uses the pre-trained VGG19 Net to extract the features of each layer of the network,separates and recombines the content diagram and multiple style diagrams,constructs a new style loss function,and fuses multiple style features to obtain a new artistic style picture. The experimental results show that the new image with multi-style fusion contains a variety of single style information,which makes the transformed image have richer visual information. The new algorithm constructs richer visual information and provides reference for new artistic creation.更多还原