摘要
传统的Image Quilting算法在进行块拼接时没有考虑到目标图像的方向场特征的变化,所以传输的纹理不具备方向性特征。针对传统算法的不足,本文提出了一种基于目标图像结构张量方向引导的改进的Image Quilting图像风格迁移算法,从而使得图像风格迁移时不仅保证了颜色及纹理特征能准确传输到目标图像中,同时通过方向场引导,确保了方向信息也能合理传输到目标图像中,同时本文运用图像内容融合技术确保目标图像的形状,边缘,内容等信息较好的留存于渲染图像中。实验表明,改进后的算法能够取得更好的图像风格迁移效果。
The traditional image quilting algorithm does not consider the change of the directional field characteristics of the target image in block splicing,so the transmitted texture does not have the directional feature.This paper proposes an improved image guided by the direction of target image structure tensor Quilting image style transfer algorithm,so that the image style transfer not only ensures that the color and texture features can be accurately transmitted to the target image,but also ensures that the direction information can be reasonably transmitted to the target image through the direction field guidance.At the same time,this paper uses image content fusion technology to ensure that the shape,edge,content and other information of the target image are well preserved in the rendered image.Experimental results show that the improved algorithm achieves better transmission stylization effect.
作者
赵杨
Zhao Yang(Animation Department,Yunnan Normal University,Kunming Yunnan,650500)
出处
《电子测试》
2021年第5期46-48,11,共4页
Electronic Test
