In recent years, Machine Vision researchers have recognized the importance
of developing algorithms from increasingly sophisticated physical models.
Such an approach holds considerable promise since the physics of image
formation links the world that must be understood to the image that is
input to a vision system. Progress has been made in modeling the properties
of surfaces, illuminants, and sensors to exploit phenomena such as
color,
shading, highlights, polarization, and interreflection for image
interpretation. The physical models have led to new algorithms for
segmenting images and recovering
properties of surfaces such as shape, spectral reflectance, and material.
References [1][2][3] are an edited collection of papers which review the
state of research in the physics-based vision subareas of color,
radiometry, and shape recovery as of 1992.
In November 1994, the Journal of the Optical Society of America A presented
a feature issue on Physics-Based Vision containing an updated review of
research in areas including sensor and reflection modeling, multispectral
processing, and multiple illumination methods. Reference [4] is the
introduction to this feature issue.
Last modified: 21 March 1996
Back to the ICVL homepage