Digital image analysis

摘要

The present invention provides a method for analyzing a digital image comprising a plurality of pixels representing a scene. The method comprising steps of obtaining pixel chromaticity information, providing, combinations each having a corresponding surface and a corresponding illuminant, performing a global scene surface-illuminant classification by determining hypothesis scores, accumulating hypothesis scores, thereby obtaining a global illuminant/surface statistic representing an estimate of a distribution of illuminants and/or surfaces in the scene as represented by the digital image. Other methods are also provided. Apparatuses for carrying out the methods are also provided.

主权项

1. A method for analyzing a digital image, the digital image comprising a plurality of pixels representing a scene, the method comprising the steps of:
obtaining pixel chromaticity information for a first set of pixels of the digital image; providing, for each surface-illuminant combination in a set of surface-illuminant combinations each having a corresponding surface and a corresponding illuminant, a corresponding reference chromaticity value representing a chromaticity corresponding to the corresponding surface of the surface-illuminant combination being illuminated by the corresponding illuminant of the surface-illuminant combination; performing a global scene surface-illuminant classification by:
determining, for each pixel of the first set of pixels, a set of one or more hypothesis scores associated with a subset of the set of surface-illuminant combinations, each hypothesis score representing a degree of correspondence between the pixel chromaticity information for said each pixel and the reference chromaticity value associated with each of the one or more surface-illuminant combinations in the subset of surface-illuminant combinations, wherein the hypothesis scores for each pixel includes a single hypothesis score for each of at least two illuminant classes having homogenous characteristics, wherein the at least two illuminant classes include a class of light sources with continuous characteristics and a class of light sources with discrete characteristics;accumulating, for illuminants and/or surfaces occurring in the surface-illuminant combinations used in determining the hypothesis scores for the first set of pixels of the digital image, the determined hypotheses scores by illuminant and/or by surface, thereby obtaining a global illuminant/surface statistic representing an estimate of a distribution of illuminants and/or surfaces in the scene as represented by the digital image.