《彩色图像的高光反射和漫射的分离》青少年教育文学精选.pdfVIP

Separation of Diffuse and Specular Reflection in Color Images Stephen Lin and Heung-Yeung Shum Microsoft Research, China fstevelin, hshumg@ Abstract made between the reflection components and two limbs in a color histogram, whose directions may be used for separa- The presence of specular reflections in images can lead tion at each pixel. A polarization approach was introduced many traditional computer vision algorithms to produce er- by Wolff [22] where polarizing filters are used to effectively roneous results. To address this problem, we propose a separate diffuse and specular reflection for areas of constant method based on the neutral interface reflection model for Fresnel ratio. Both color and polarization are employed by separating the diffuse and specular reflection components Nayar et al. [16] to constrain estimates of reflection com- in color images. From two photometric images without cal- ponents, and Sato and Ikeuchi [18] achieve separation using ibrated lighting, the illuminant chromaticity is estimated, color signatures generated from many images taken with a and the RGB intensities of the two reflection components moving light source. are computed for each pixel using a linear model of surface These previous methods produce successful results un- reflectance. Unlike most previous methods, the presented der conditions prescribed by their respective approaches. A technique does not assume any dependencies among pixels, difficulty of polarization-based procedures is not only the such as regionally-uniform surface reflectance.