Shape from Shading using Color Segmentation Based Depth Adjustment

In order to create a good 3D reconstruction from an image, shape information of the object of interest is needed. One way to acquire it is based on the shading information. This can be solved using the Shape from Shading method. Unfortunately this method has some drawbacks, namely the depth ambiguity problems. Two points which have the same intensity will have the same depth, although in reality they have different depth. Color also plays an important role. Some color generates higher intensity than the others. This makes the resulting output of the system to reconstruct the object incorrectly. We propose a color segmentation based depth adjustment to overcome this problem. First we cluster the image into color clusters. Based on this color cluster we will adjust the depth of vertices belonging to the higher intensity cluster accordingly. The proposed method can reduce the effect of the depth ambiguities.