Kinect quality enhancement for triangular mesh reconstruction with a medical image application

Abstract

This chapter presents a method to estimate proportion between skin burn area and body surface using computer vision techniques. The data are derived using Microsoft Kinect. We first show a comprehensive Kinect calibration method. Then the color image is segmented into 3 sections, background, burn area, and body area. The segmentation method developed is based on watershed algorithm and Chebyshev’s inequality. The segmented color image is to be mapped with the depth image to generate triangular mesh. We discover that reconstructing the 3D mesh, using marching cube algorithm, directly from Kinect depth information is erroneous. Hence, we propose to filter the depth image first. After the enhanced mesh is derived, the proportion between 3D meshes of burn area and 3D meshes of body surface can be found using Heron’s formula. Finally, our paradigm is tested on real burn patients.