3D Point Cloud Face Reconstruction From 2d Image Using Pixel Density Technique
3D face reconstruction has been broadly studied and used in many fields. The aim of this project is to construct a three dimensional (3D) point cloud face from two dimensional (2D) images. When a 2D image is uploaded into the system, Face detection function detects the face from the 2D image. Then,...
Saved in:
Main Author: | |
---|---|
Format: | Academic Exercise |
Language: | English |
Published: |
2014
|
Subjects: | |
Online Access: | https://eprints.ums.edu.my/id/eprint/17529/1/3D%20Point%20Cloud%20Face%20Reconstruction%20From%202d%20Image%20Using%20Pixel%20Density%20Technique.pdf https://eprints.ums.edu.my/id/eprint/17529/ |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | 3D face reconstruction has been broadly studied and used in many fields. The aim of this project is to construct a three dimensional (3D) point cloud face from two dimensional (2D) images. When a 2D image is uploaded into the system, Face
detection function detects the face from the 2D image. Then, the Linear Pixel Shuffling technique is used to extract the pixel from the image. After that, the depth data is calculated. In this system, GUI menu is used so that users can interact with the system. There are other alternative choices to allow the user to capture a new photo instead using the only default photo. The basic transformation like, rotation, cell size and scale is called in the GUI menu. There are three experiments conducted to test the accuracy of the 3D model. The first test is the brightness test. During the test, three same images but varying in the level of brightness is used to compare the differences in the 3D reconstruction of the face. Through this experiment, sufficient amounts of brightness such as < 40%, 40%-60% and > 60% are needed to get an accurate result. The cell size test is carried out in the second experiment. In this test, as the cell size exceeds 6, the accuracy of the 3D model becomes lower. Furthermore, the shape of the 3D model will become cubic-look and unclear. Lastly, the memory test is examined to check the efficiency of the system. The memory consumption is low, 30-60 MB because there are not much complicated a lgorithm present in the system.In the future work, the 3D face model not only can be done by reconstructing the 3D face but also other objects. Besides, a more realistic 3D face model can be constructed instead of only construct in point cloud form. |
---|