Conference Paper

Physics-Based Extraction of Intrinsic Images from a Single Image.

DOI: 10.1109/ICPR.2004.695 Conference: Pattern Recognition, 2004. ICPR 2004. Proceedings of the 17th International Conference on, Volume: 4
Source: DBLP

ABSTRACT A technique for extracting intrinsic images, including the reflectance and illumination images, from a single color image is presented. The technique first convolves the input image with a prescribed set of derivative filters. The pixels of filtered images are then classified into reflectance-related or illumination-related based on a set of chromatic characteristics of pixels calculated from the input image. Chromatic characteristics of pixels are defined by a photometric reflectance model based on the Kubelka-Munk color theory. From the classification results of the filtered images, the intrinsic images of the input image can be computed. Real images have been utilized in our experiments. The results have indicated that the proposed technique can effectively extract the intrinsic images from a single image.

0 Followers
 · 
149 Views
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In computer vision, research on reconstructing geometric shape of physical objects by machines has been one of primary issues, and numerous methods for obtaining range information from images have been presented. Active range sensing using structured light based on triangulation is the most accurate and reliable method for obtaining 3D geometric information. Although the geometry of objects varies with time in most cases most previous structured-light methods have assumed that the object scene is static or near static. Moreover, the modeling of dynamic object has hardly been implemented even by the rare structured-light methods for range sensing of dynamic objects. For reliable range sensing of dynamic objects, range information should be acquired from one or slightly more video frames and the acquired range information should be insensitive to object color and ambient illumination. This dissertation presents color structured-light for rapid high-resolution range imaging which applicable to the modeling of dynamic objects. We propose color-phase processing methods to implement single-frame range imaging with color sinusoidal light. We analyze color-phase distortion for estimating accurate phase information, and propose a reliable unwrapping method for imaging with high-frequency sinusoidal light. In addition, we present color-stripe structured-light methods for single-frame or double-frame imaging based on novel color analysis, permutation overlapping and sign codification which maximizes acquisition speed, range resolution and accuracy. The imaging area of a 3D object using a single projector is restricted since the structured light is projected only onto a limited area of the object surface. Employing additional projectors to broaden the imaging area is a challenging problem since simultaneous projection of multiple patterns results in their superposition in the light-intersected areas and the recognition of original patterns is by no means trivial. This dissertation presents a method of multi-projector color structured-light vision based on projector-camera triangulation. By analyzing the behavior of superposed-light colors in a chromaticity domain we show that the original light colors cannot be properly extracted by the conventional direct estimation. We disambiguate multiple projectors by multiplexing the orientation of projector pattern so that the superposed patterns can be separated by explicit derivative computations. Experimental studies are carried out to demonstrate the validity of the presented method. The proposed techniques can be applied to high-resolution motion capture for non-rigid dynamics, and to virtual creation of aesthetic scenes and animations in both of photorealistic and non-photorealistic senses.
    Department of Electronic Engineering, Sogang University, , Degree: PhD
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In computer vision, research on reconstructing geometric shape of physical objects by machines has been one of primary issues, and numerous methods for obtaining range information from images have been presented. Active range sensing using structured light based on triangulation is the most accurate and reliable method for obtaining 3D geometric information. Although the geometry of objects varies with time in most cases most previous structured-light methods have assumed that the object scene is static or near static. Moreover, the modeling of dynamic object has hardly been implemented even by the rare structured-light methods for range sensing of dynamic objects. For reliable range sensing of dynamic objects, range information should be acquired from one or slightly more video frames and the acquired range information should be insensitive to object color and ambient illumination. This dissertation presents color structured-light for rapid high-resolution range imaging which applicable to the modeling of dynamic objects. We propose color-phase processing methods to implement single-frame range imaging with color sinusoidal light. We analyze color-phase distortion for estimating accurate phase information, and propose a reliable unwrapping method for imaging with high-frequency sinusoidal light. In addition, we present color-stripe structured-light methods for single-frame or double-frame imaging based on novel color analysis, permutation overlapping and sign codification which maximizes acquisition speed, range resolution and accuracy. The imaging area of a 3D object using a single projector is restricted since the structured light is projected only onto a limited area of the object surface. Employing additional projectors to broaden the imaging area is a challenging problem since simultaneous projection of multiple patterns results in their superposition in the light-intersected areas and the recognition of original patterns is by no means trivial. This dissertation presents a method of multi-projector color structured-light vision based on projector-camera triangulation. By analyzing the behavior of superposed-light colors in a chromaticity domain we show that the original light colors cannot be properly extracted by the conventional direct estimation. We disambiguate multiple projectors by multiplexing the orientation of projector pattern so that the superposed patterns can be separated by explicit derivative computations. Experimental studies are carried out to demonstrate the validity of the presented method. The proposed techniques can be applied to high-resolution motion capture for non-rigid dynamics, and to virtual creation of aesthetic scenes and animations in both of photorealistic and non-photorealistic senses.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Intrinsic images, including reflectance and illumination images, are desirable to many vision applications. An improved method for extracting intrinsic images from a single color image with integrated measures is presented. To start with, the input image convolves with a predefined set of derivative filters. The pixels of filtered images are then classified into reflectance-related or illumination-related using a criterion measure comprising three measures of filtered pixels calculated from the input image. The three measures are denoted as chromatic measure, blur measure, and intensity measure. Finally, the intrinsic images of the input image can be computed from the classification results of the filtered images. Both synthetic and real images have been utilized in our experiments. The results demonstrated that the proposed technique can effectively extract the intrinsic images from a single image.
    IASTED International Conference on Artificial Intelligence and Applications, part of the 23rd Multi-Conference on Applied Informatics, Innsbruck, Austria, February 14-16, 2005; 01/2005
Show more