3D Wavelet Transform-based 3D Shape Recovery Using Noisy Images

This paper proposes a shape from focus method that uses 3D discrete wavelet transform-based focus measure operator. This operator decomposes a noisy input image sequence in approximation and detail components up to certain levels. At each level, an image focus volume is extracted from the energy of detail components. Different-scaled structural information of the object in the scene is contained by these focus volumes. Then, 3D weighted least squares is applied that aggregates these focus volumes and a single focus volume is obtained. This across-the-scale aggregation leads to a single image focus volume that embodies all the important features which are required for the accurate 3D shape recovery. Finally, the depth map is recovered from the resultant image focus volume using conventional approach. Experiments have been performed using image sequences of synthetic and real objects. Experimental results demonstrate that the proposed method is effective for recovering 3D shape under noisy condition.