3-D texture model of natural scene image

A novel 3-D texture model is developed by considering the scene image as the superposition of a random texture image and a smooth shaded image. This smoothed surface is obtained by taking the average of pixel intensities from the natural image. A texture image can be generated by subtracting the smoothed image from the original one. Then, the shape-from-shading technique is applied to estimate the orientation of the surface. For describing a random texture image, the long-correlation periodic model is chosen because of its good performance in texture synthesis and its ability to represent the coarseness and the pattern of the surface at the same time. Then, a projection technique is developed to deal with a random model which has a nonisotropically distributed texture pattern. For estimating the parameters, a hybrid method which uses both the least-square and the maximum-likelihood estimates is applied directly to the given intensity function.