Remove specular highlight from facial images using deep learning
In this paper, I present an image pre-processing technique, known as specular highlight removal in facial images. In comparison to previous works that depend on physical and statistical properties of human skin and faces, I propose a deep learning algorithm using autoencoder to generate highlight-fr...
Saved in:
| 主要作者: | |
|---|---|
| 其他作者: | |
| 格式: | Final Year Project |
| 語言: | English |
| 出版: |
Nanyang Technological University
2021
|
| 主題: | |
| 在線閱讀: | https://hdl.handle.net/10356/148186 |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 機構: | Nanyang Technological University |
| 語言: | English |
| 總結: | In this paper, I present an image pre-processing technique, known as specular highlight removal in facial images. In comparison to previous works that depend on physical and statistical properties of human skin and faces, I propose a deep learning algorithm using autoencoder to generate highlight-free/diffuse facial images. By taking facial images with specular highlight as the input, the autoencoder model that has been trained on the corresponding diffuse images will attempt to reconstruct diffuse images from the features that it has learned. The results obtained in this project demonstrate a good learning efficiency and reliable reconstruction of the facial image. I have also built a dataset of synthetic facial images in a different lighting environment that was used in this facial highlight removal experiment. |
|---|