An approach to color constancy using multiple images

Tsukada, M.; Ohta, Yasuyuki

doi:10.1109/iccv.1990.139557

Cited by 29 publications

(14 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Existing approaches have mainly focused on the problem of color constancy, where the goal was to extract surface color, i.e., surface reflectance properties of objects, in order to obtain a stable perception of the color of an object under varying illumination conditions. As this problem is under-constrained, most methods make some assumptions about either the surface being imaged [23], or about the illumination conditions [25,14,11,32], or both [10]. Other approaches also exist that try to recover image color, i.e., the color of the objects as they appear under the present illumination conditions, accounting separately for artifacts such as specularities on shiny surfaces [22].…”

Section: Surface Color Image Color Perceptual Colormentioning

confidence: 99%

Data and model-driven selection using color regions

Syeda-Mahmood¹

1992

Computer Vision — ECCV'92

View full text Add to dashboard Cite

A L, 2 0 1 1 1 o reuaxeo tQ a.erqe 'i.~ o r~ . -0,3 r-e time ?or reviewing instructions searcn nq ex -t3ca f.. Distribution of this document is unlimited ABSTRACT (Maximum 200 words)A key problem in model-based object recognition is selection, namely, the problem of determining which regions in the image are likely to come from a single object. In this paper we present an approach that uses color as a cue to perform selection either based solely on imagedata (data-driven), or based on the knowledge of the color description of the model (model-driven). Specifically, the paper argues for the specification of color in terms of color categories as being appropriate for the task of selection. These color categories are used to develop a fast region segmentation algorithm that extracts perceptual color regions in images. The color regions extracted form the basis for performing data and model-driven selection. Data-driven selection is achieved by (continued on back) selecting salient color regions as judged by a color-saliency measure that emphasizes attributes that are also important in human color perception. The approach to model-driven selection, on the other hand, exploits the color region information in the model to locate instances of the model in a given image. The approach presented tolerates some of the problems of occlusion, pose and illumination changes that make a model instance in an image appear different from its original description. Finally, the utility of color-based data and model-driven selection is discussed in the context of reducing the search involved in recognition. A key problem in model-based object recognition is selection, namely, the problem of determining which regions in the image are likely to come from a single object. In this paper we present an approach that uses color as a cue to perform selection either based solely on imagedata (data-driven), or based on the knowledge of the color description of the model (model-driven). Specifically, the paper argues for the specification of color in terms of color categories as being appropriate for the task of selection. These color categories are used to develop a fast region segmentation algorithm that extracts perceptual color regions in images. The color regions extracted form the basis for performing data and model-driven selection. Data-driven selection is achieved by selecting salient color regions as judged by a color-saliency measure MomVi that emphasizes attributes that are also important in human color per- 0Eception. The approach to model-driven selection, on the other hand, exploits the color region information in the model to locate instances of the model in a given image. The approach presented tolerates some so• of the problems of occlusion, pose aad illumination changes that make a model instance in an image appear different from its original descrip-O'4 30 tion. Finally, the utility of color-based data and model-driven selection is discussed in the context of reducing the search involved in recognition. 93 1 27 11 SELECTIO...

show abstract

Section: Surface Color Image Color Perceptual Colormentioning

confidence: 99%

Data and model-driven selection using color regions

Syeda-Mahmood¹

1992

Computer Vision — ECCV'92

View full text Add to dashboard Cite

show abstract

“…al. [16] have shown that a difference in illumination, once it has been identified, provides additional constraints that can be exploited to obtain colour constancy, but they do not provide an automatic method of determining when such a difference exists. We present a new algorithm that first uncovers the illumination variation in an image and then uses the additional constraint it provides to obtain better colour constancy.…”

Section: Introductionmentioning

confidence: 99%

Color Constancy for Scenes with Varying Illumination

Barnard

Finlayson

Funt

1997

Computer Vision and Image Understanding

View full text Add to dashboard Cite

Abstract. We present an algorithm which uses information from both surface reflectance and illumination variation to solve for colour constancy. Most colour constancy algorithms assume that the illumination across a scene is constant, but this is very often not valid for real images. The method presented in this work identifies and removes the illumination variation, and in addition uses the variation to constrain the solution. The constraint is applied conjunctively to constraints found from surface reflectances. Thus the algorithm can provide good colour constancy when there is sufficient variation in surface reflectances, or sufficient illumination variation, or a combination of both. We present the results of running the algorithm on several real scenes, and the results are very encouraging.

show abstract

“…The bilinear relationship (3) has long been observed in the computer vision literature [1,23,22]. The important difference here is that our representation is independent of the basis function selection.…”

Section: Color Image Formation Modelmentioning

confidence: 99%

“…The important difference here is that our representation is independent of the basis function selection. In [1,22] light spectrum basis functions are obtained by extracting principle components from a large set of light samples measured by spectroradiometers, and camera sensitivity functions are selected heuristically, for example to approximate the human eye response [19]. Such choices work fine for simulations.…”

Section: Color Image Formation Modelmentioning

confidence: 99%

“…The generalized diagonal transform [5] performs better than the diagonal transformation model, but assumes very low dimensionality of the reflectance and spectrum. Finite dimensional linear models [12,1,22] require explicit use of illumination basis functions and camera sensitivity functions, which are not always available and accurate. Our observation is that only the coefficients used to combine basis functions are important for color constancy.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation