Non-pre-process calibration of depth image based on fuzzy c-mean

In this paper, a non-preprocess calibration of depth image is proposed Take advantage of FCM to acquire the depth value distribution in the depth image. After that, according to relation among all the centroids of cluster, the real distance is estimated. Then, the error of the depth value is able to be compensated. When utilize the proposed method, plenty of pre-process for calibration can be avoided, such as using chessboard to capture the camera parameters, or recording measurement error in advance. Therefore, time cost, inconvenient, and human error for calibration can be reduced significantly. Utilize the proposed method can offer the users a reliable depth camera without traditional calibration procedure. At last, the proposed method is verified by comparing the consequents with traditional depth calibration and laser rangefinder. The results show it has an outstanding performance.