Jacobus J Barnard
We develop and demonstrate an object recognition system capable of accurately detecting, localizing, and recovering the kinematic configuration of textured animals in real images. We build a deformation model of shape automatically from videos of animals and an appearance model of texture from a labeled collection of animal images, and combine the two models automatically. We develop a simple texture descriptor that outperforms the state of the art. We test our animal models on two datasets; images taken by professional photographers from the Corel collection, and assorted images from the web returned by Google. We demonstrate quite good performance on both datasets. Comparing our results with simple baselines, we show that for the Google set, we can recognize objects from a collection demonstrably hard for object recognition. © 2005 IEEE.
This venue is a peer reviewed, competitive conference (acceptance rate: 26%) and the full paper is published as part of the conference proceedings [ CSRanking endorsed, A* ]
Sensor sharpening [J. Opt. Soc. Am. A 11, 1553 (1994)] has been proposed as a method for improving computational color constancy, but it has not been thoroughly tested in practice with existing color constancy algorithms. In this paper we study sensor sharpening in the context of viable color constancy processing, both theoretically and empirically, and on four different cameras. Our experimental findings lead us to propose a new sharpening method that optimizes an objective function that includes terms that minimize negative sensor responses as well as the sharpening error for multiple illuminants instead of a single illuminant. Further experiments suggest that this method is more effective for use with several known color constancy algorithms. © 2001 Optical Society of America.
We present an extensive data set for color research that has been made available online (www.cs.sfu.ca/̃colour/data). The data are especially germane to research into computational color constancy, but we have also aimed to make the data as general as possible, and we anticipate a wide range of benefits to research into computational color science and computer vision. Because data are useful only in context, we provide the details of the collection process, including the camera characterization, and the data used to determine that characterization. The most significant part of the data is 743 images of scenes taken under a carefully chosen set of 11 illuminants. The data set also has several standardized sets of spectra for synthetic data experiments, including some data for fluorescent surfaces. © 2002 Wiley Periodicals, Inc. Col. Res. Appl.
Color images often must be color balanced to remove unwanted color casts. We extend previous work on using a neural network for illumination, or white-point, estimation from the case of calibrated images to that of uncalibrated images of unknown origin. The results show that the chromaticity of the ambient illumination can be estimated with an average CIE Lab error of 5ΔE. Comparisons are made to the grayworld and white patch methods.