Acquisition and Pre-processing of 3D Facial Image
Wang Jun-jua · Xingshi jishu · 2015
Three-dimensional(3D) facial image provides diversified information and be comes more and more important. However, acquiring 3D facial images is a challenge especially when we need to get a large number of high-resolution images. It includes device selecting, quality controlling, data pre-processing and so on. In the past three months, over 1100 3D facial images had been captured for our project. This paper proposed some technique improvements when capturing 3D facial images. The first one was device selecting. Artec Spider portable 3D scanner was selected based on an extensive performance comparison. The scanner could capture depth informat ion and color texture. We could get high-resolution 3D image with scan resolution of 0.1mm. The second one was to control the image quality when we captured a large number of samples. Due to the portability of the scanner, the capturing results could be affected by the operator. To avoid this problem, we designed a fixed scanning path to get more stable and reliable results. The third one was data pre-processing. The original scanning result was a combination of curved surfaces, which were aligned by auto-registration. We proposed some manual interve ntion methods to enhance the c omposed precision. The final one was the efficiency. It was important to sample a large scale region with reliable data and relatively short time. We analyzed the relation of pre-processing time, image quality and the number(fra me) of curved surfaces, and concluded that the optimum number of sample frame was 500~700.