The present invention relates to a depth map correction method which comprises the steps of: calculating a geometric transformation function between a current frame and predetermined neighboring frames before a current frame; projecting the neighboring frames onto the current frame by using the calculated geometric transformation function; and correcting a depth map of the current frame by calculating the average of a depthv value in a pixel of the current frame and a pixel of the neighboring frames corresponding to the pixel of the current frame. Therefore, an effect similar to maintaining a three-dimensional model can be generated by calculating a three-dimensional relation for serveral depth maps to use the three-dimensional relation for correcting the depth map, and a cumulative error which is recognized as a big problem in all camera trajectory esetimation method can be reduced by reducing the cumulative speed of the error by setting a reference frame.
展开▼