Illumination-Aware Age Progression

基于光照感知的年龄增长技术

作者使用了一种全新的技术，在基于数万张同一年龄段人群照片的基础上，构建出“年龄模板”。通过对input的儿童或青少年照片进行处理，做出年龄预测。

 

What will a child look like in 20 years?

 

文章中作者告诉我们：The transformation from child to adult is dominated by tcraniofacial growth, in which the forehead slopes backward, the head expands, and the lower portion of the face extends downward [8]. Changes in later years are dominated by growth of the nose, narrowing of the eyes, the formation of wrinkles and other textural changes.可以得知，从儿童到成人这个过程，额头，眼睛的大小，鼻子的长短，脸颊的形状都会发生变化。

 

作者借鉴的方法和成果有：

• Galton [10] generated average images by taking several exposures of portraits on the same photographic plate.
• Bensen and Perrett [2] showed that dramatically better composites can be obtained by first aligning facial features (208 fiducials) and warping the images to a reference prior to averaging.

基于此，在这个问题上，作者提出在基于数万张同一年龄段人群照片的基础上，构建出“年龄模板”。

 

 

 

对比Before flow和After flow后，我们可以发现经过处理的照片更清晰。但是它处于一种理想情况既完美光照的情况下。因此当我们做Re-illuminated处理，得到更真实更符合现实的图片。

 

Relighting的关键优点如下：Two key advantages of relighting are that 1) it generates a more realistic set of average images, bringing out fine details that are only visible with proper shading, and 2) we can align the lighting across the set of averages, to enable comparing changes at different ages.

 

 

 

一个完整的基于光照年龄增长处理为

 

 

总结：这篇文章暴露出自己对图形图像方面知识的缺陷。