基于多尺度分析的自动人脸照片移植
针对人脸移植中输入图像与目标图像的脸部姿态、光照环境与颜色分布不一致的问题，提出了一种基于多尺度分析的自动人脸照片移植方法。通过多线性模型从单张图像中恢复三维人脸模型，从而自动变换输入图像中的人脸姿态。提出了一种多尺度增强与融合算法，根据目标图像的细节特征对输入图像自动调整，并通过无缝融合合成新的人脸照片。实验结果表明，该方法可以使输入图像有效匹配目标图像的明暗变化与颜色分布，并自适应调整局部细节。该方法对各种人脸图像之间的移植鲁棒性高，合成照片真实感强。...
Saved in:
| Published in | 计算机应用研究 Vol. 34; no. 11; pp. 3510 - 3513 |
|---|---|
| Main Author | |
| Format | Journal Article |
| Language | Chinese |
| Published |
西北工业大学电子信息学院,西安,710129
2017
|
| Subjects | |
| Online Access | Get full text |
| ISSN | 1001-3695 |
| DOI | 10.3969/j.issn.1001-3695.2017.11.069 |
Cover
| Summary: | 针对人脸移植中输入图像与目标图像的脸部姿态、光照环境与颜色分布不一致的问题，提出了一种基于多尺度分析的自动人脸照片移植方法。通过多线性模型从单张图像中恢复三维人脸模型，从而自动变换输入图像中的人脸姿态。提出了一种多尺度增强与融合算法，根据目标图像的细节特征对输入图像自动调整，并通过无缝融合合成新的人脸照片。实验结果表明，该方法可以使输入图像有效匹配目标图像的明暗变化与颜色分布，并自适应调整局部细节。该方法对各种人脸图像之间的移植鲁棒性高，合成照片真实感强。 |
|---|---|
| Bibliography: | 51-1196/TP The major challenge in the face photo transfer is that the source and target images have various face poses, lighting conditions and color distributions. This paper proposed a face transfer method based on multi-scale analysis. It used the multi-linear model to reconstruct 3D models for the source and target faces, and then it automatically transformed the source image. It developed a new multi-scale image synthesizing algorithm, which could adapt the source image details to the target and smoothly blend images. The experimental result shows that this method can effectively match the shading and the color of the target image, and also automatically adjust the local details. The synthesizing photos exhibit normal and realistic faces, without any facial artifacts. The final results show that this method is robust for various facial images face transfer; multi-linear model; multi-scale analysis Huang Yanhui, Fan Yangyu, Su Xuhui （ School of Electronics ＆ Information, Northwestern Polytechnical University |
| ISSN: | 1001-3695 |
| DOI: | 10.3969/j.issn.1001-3695.2017.11.069 |