Modulating Energy Distribution of Reflected Light Based on Images

Abstract

Based on IBMR techniques, this paper study energy distribution of reflection light from object surface and a method is presented to separate elements of specular and diffuse reflection in an image. The ray projected from object surface to images includes diffuse and specular energy distribution. The ray projected from object surface to images includes energy distribution from ambient light, diffuse and specular light. The ray intensity of diffuse reflection or ambient light is irrelevant to viewpoint, but that of specular reflection is relevant to viewpoint and will change with the moving of viewpoint. This paper suppose that the minimal energy of reflection lighting in n corresponding points is energy brought by diffuse reflection and ambient light, so diffuse and specular reflection element are separated.

The separated energy distribution of diffuse light depicts the reflection ability and weight of colors, and that of specular propagation depict the brightness of object surface. Appointed new specular and diffuse ratio to the separated images, scene images with different reflectance properties of the objects surfaces can be obtained, and virtual images with different ratio of energy distribution from specular and diffuse reflection can be reconstructed, furthermore it can be achieve to change the object colors and brightness only based on images.

When rendering images, characteristic of object surface are often adjusted to acquire satisfying visual impression. IBMR techniques have recently received much attention as a powerful alternative to traditional geometry-based techniques for image synthesis. Based on IBMR, this paper present a method to separate energy distribution of diffuse and specular reflection in an image. If re-distributing ratio to the separated images, then the surface characteristic of object in image can be changed, and different impression of images can be obtained.

In the method, inputs are only a few calibrated images, algorithm is simple. Object surface reflection character, geometric model of the scene and the position of lighting are not needed.