Acceleration of the reflectance field acquisition using independent component analysis

In order to generate photorealistic images, a central problem in computer graphics is the description of an object reflectance model. The reflectance field technique describes the object surface properties and can be used for photorealistic rendering. The reflection of surfaces can be described as a high dimensional reflectance function. For complex surfaces, an analytical model is not always easy to formulate, therefore the direct real-world surface acquisition is preferred. The reflectance is typically acquired with a camera or array of cameras that capture the reflectance field of the object surface but the reflectance information can be composed of thousands of images, depending on the surface material properties and the camera resolution. In this work we proposes a systematic strategy that incorporates Independent Component Analysis (ICA) to acquire the reflectance field and reducing by orders of magnitude the required number of captured images and keeping the same reflectance field quality. In our experiments, a reflectance field can be obtained with only 26 images, compared to the classical approach that require thousands of images, with an error less than 2%.