Colour Texture Fusion of Range Images

Alexander Agathos

When constructing complete 3D models, the fusion of different colour images taken with a CCD requires several aspects to be considered. This is in addition to the 3D shape fusion arising from merging multiple 3D scans. First a calibration of the camera with the range data is needed: for each point of the range image a pixel (texel) of the picture corresponding to this view should be found. For this kind of calibration various approaches have been proposed (eg [1]). When reconstructing the texture of a manifold reconstructed from multiple range images, a 3D point could be observed in more than one intensity image. This means that the manifold may be coloured by patches from different images. This can lead to an unrealistic representation of the object’s texture because of :

· Ghosting effects (images not calibrated correctly with the range views)

· Colour differences between the images, an issue that has to do with illumination and camera quality

Assuming that we have calibrated correctly the images with the range views we are going to deal with the colour differences between the images.

A model showing the fusion of two images belonging to two different views of a model. This is before correction. Note that there are slightly brownish and alternatively greyer patches on the nose and elsewhere.

After processing, the color transitions between source views are not visible and the hippo has a more uniform appearance.

Our approach [12] in correcting the color of the images acquired from two different views has two stages:
1) A global correction that does a colour transform estimated from the two different colours observed at pixels in the overlap region
2) A local correction to smooth out small colour variation boundaries that are left out.

Some results of our colour correction scheme can be seen here.

References:

[1] C. Robertson, and R. B. Fisher, “Empirical Calibration Method for Adding Colour to Range Images”, 3D Data Processing, Visualization and Transmission, Int. Conf., Padova, 2002, pp. 558-561

[2] F. Bernardini, I. Martin, and H. Rushmeier, “High-quality Texture Reconstruction from Multiple Scans”, IEEE Transactions on Visualization and Computer Graphics, 7(4), 2001, pp. 318-332

[3] M. Callieri, P. Cignoni, C. Rocchini, and R. Scopigno, “Weaver, an automatic texture builder”, 3D Data Processing, Visualization and Transmission, Int. Conf., Padova, 2002, pp. 562-565

[4] C. Rocchini, P. Cignoni, C. Montani, and R. Scopigno, “Acquiring, Stitching and Blending Appearance Attributes on 3D Models”, The Visual Computer, Springer International, 18(3), 2002, pp. 186-204

[5] M. Soucy, G. Godin, R. Baribeau, F. Blais, and M. Rioux, “Sensors and Algorithms for the Constructionof Digital 3-D Colour Models of Real Objects”, ICIP Proceedings, Lausanne, Switzerland, 1996, pp. 409-412

[6] A. Jain, and A. Vailay, “Image Retrieval Using Colour and Shape”, Pattern Recog., 29(8), 1996, pp. 1233-1244

[7] P. Besl, and N. McKay, “A Method for Registration of 3-D Shapes”, Trans. PAMI, 14(2), 1992, pp. 239-256

[8] Y. Chen, and G. Medioni, “Object Modeling by Registration of Multiple Range Images”, Proc. IEEE Conf. on Robotics and Automation, 1991, pp. 145-155

[9] S. Rusinkiewicz, and M. Levoy, “Efficient Variants of the ICP Algorithm”, Proc. 3DIM, 1997, pp. 145-152

[10] E. Reinhard, M. Ashikhmin, B. Gooch, P. Shirley, "Color Transfer Between Images", IEEE Computer Graphics and Applications, 21(5), (2001)(special issue on Applied Perception), pp. 34-41

[11] H. Rushmeier, F. Bernardini, “Computing consistent normals and colors from photometric data”, Proc. 3DIM, 1999, pp. 99-108

[12] A. Agathos, R. B. Fisher, "Colour Texture Fusion of Multiple Range Images", Proc. 4th Int. Conf. on 3-D Digital Imaging and Modeling, Banff, to appear, 2003.