Accuracy of depth and orientation estimation based on texture characteristics

We consider the problem of estimating the depth and orientation (geometric parameters) of a flat surface that is viewed from multiple cameras. It is well known that the ability to obtain accurate geometric parameter estimates depend on the presence of sharp features in the texture of the imaged object. We quantify this statement by formulating the problem as an estimation problem and then applying tools from estimation theory. Our analysis results in closed-form expressions for the accuracy of surface geometry parameter estimation in terms of the object´s texture, orientation and depth, and in terms of the number of views. The theoretical results obtained are validated experimentally.