Cornell Box [Pen98a]
Cornell University Program of Computer Graphics
Cornell Seal

The Color Histogram and its Applications in Digital Photography.

Liang Peng.

PhD thesis, Cornell University, 1998.

The color histogram is a tool that has been widely used in many image processing and computer vision applications such as image retrieval and image segmentation. In these applications, the features in the color histogram are used as signatures for feature comparison, scene analysis and object recognition. Various histogram features were explained previously by several models in computer vision which incorporate the physical process of light reflection. However, most of these approaches are mainly based on a dichromatic reflectance model, and only consider the lighting coming directly from the light source in the scene. In this dissertation, we present a comprehensive study of the formation of color histogram structures based on both the theory and simulation of light reflection and transport in the three dimensional world with the full spectrum of visible light by using advanced computer graphic techniques. By using a physically accurate, precisely controlled computer graphic environment, we have been able to successfully isolate and reproduce the features in the color histogram observed by previous researchers. Furthermore, we have also produced more complex new features, such as off-plane thickening and ``banana'' shaped structures in the color histogram by incorporating global illumination in our image synthesis procedure. Several physically based reflectance models are used to illustrate and explain the formation of these features in the color histogram. We then use the color histogram analysis techniques in a digital photographic application to extract the spectral information about the illumination in the scene and spectral information about the diffuse component of the reflectance of a dichromatic surface. This spectral information is then used to predict the appearance of objects in the image under different illumination spectra. Our method is demonstrated to show improved performance over some traditional illumination compensation methods, and has great potential in digital photography.

Last updated 07/22/02 PCG www Home