Improved methods of estimating shape from shading using the light source coordinate system
Artificial Intelligence
A New Sense for Depth of Field
IEEE Transactions on Pattern Analysis and Machine Intelligence
Shape from shading
Height and gradient from shading
International Journal of Computer Vision
Surface shape and curvature scales
Image and Vision Computing
Parametric Shape-from-Shading by Radial Basis Functions
IEEE Transactions on Pattern Analysis and Machine Intelligence
IEEE Transactions on Pattern Analysis and Machine Intelligence
Remote Sensing: Digital Image Analysis
Remote Sensing: Digital Image Analysis
Densification of Digital Terrain Elevations Using Shape from Shading with Single Satellite Imagery
ICCS '01 Proceedings of the International Conference on Computational Science-Part II
ICCS '02 Proceedings of the International Conference on Computational Science-Part III
Novel Shape-From-Shading Methodology with Specular Reflectance Using Wavelet Networks
ISNN '07 Proceedings of the 4th international symposium on Neural Networks: Part II--Advances in Neural Networks
Hi-index | 0.00 |
Digital terrain models (DTMs) in the present context are simply regular grids of elevation measurements over the land surface. DTMs are mainly extracted by applying the technique of stereo measurements to images available from aerial photography and/or remote sensing. Enormous amounts of local and global DTM data with different specifications are now available. However, numerous geoscience and engineering applications need denser and more accurate DTM data. Collecting additional height data in the field, if not impossible, is either expensive or time consuming or both. Stereo aerial or satellite imagery is often unavailable and very expensive to acquire. Interpolation techniques are fast and cheap, but have their own inherent difficulties and problems, especially in rough terrain. Advanced space technology has provided much single (if not stereo) high-resolution satellite images almost worldwide. Besides, shape from shading (SFS) is one of the methods to derive the geometric information about the objects from the analysis of the monocular images. This paper discusses the idea of using the SFS method with single high resolution imagery to optimize the interpolation techniques used in densifying regular grids of heights. Three different methodologies are briefly explained and then implemented with both simulated and real data. Numerical results are briefly discussed and a short discussion on how to make the computations more efficient will be presented. The implemented algorithms show that one can easily take advantage of parallel processing techniques to speed up the highly demanding computations involved in this application. Finally, a few remarks and conclusions are drawn.