Illumination problem

About: Illumination problem is a research topic. Over the lifetime, 93 publications have been published within this topic receiving 5859 citations.

Logarithm Gradient Histogram: A general illumination invariant descriptor for face recognition

Abstract: In the last decade, illumination problem has been the bottleneck of robust face recognition system. Extracting illumination invariant features becomes more and more significant to solve this issue. However, existing works in this field only consider the variations caused by lighting direction or magnitude (denoted as homogeneous lighting), while the spectral wavelength is always ignored in most of the developed illumination invariant descriptors. In this paper, we claim that the spectral wavelength is important, and we propose a novel gradient based descriptor, namely Logarithm Gradient Histogram (LGH), which takes the illumination direction, magnitude and even the spectral wavelength together into consideration (denoted as heterogeneous lighting). Our proposal contributes in the following three-folds: (1) we incorporate homogeneous filtering to alleviate the illumination effect for each image and extract two illumination invariant components, namely logarithm gradient orientation (LGO) and logarithm gradient magnitude (LGM); (2) we propose an effective postprocessing strategy to guarantee the fault-tolerant ability and generate a histogram representation to integrate both LGO and LGM; (3) we present thorough theoretical analysis on the illumination invariant properties for our proposed method. Experimental results on CMU-PIE, Extended YaleB and HFB databases are reported to verify the effectiveness of our proposed method.

Wavelet algorithms for illumination computations

Abstract: One of the core problems of computer graphics is the computation of the equilibrium distribution of light in a scene. This distribution is given as the solution to a Fredholm integral equation of the second kind involving an integral over all surfaces in the scene. In the general case such solutions can only be numerically approximated, and are generally costly to compute, due to the geometric complexity of typical computer graphics scenes. For this computation both Monte Carlo and finite element techniques (or hybrid approaches) are typically used. A simplified version of the illumination problem is known as radiosity, which assumes that all surfaces are diffuse reflectors. For this case hierarchical techniques, first introduced by Hanrahan et al. (32), have recently gained prominence. The hierarchical approaches lead to an asymptotic improvement when only finite precision is required. The resulting algorithms have cost proportional to $O(k\sp2 + n)$ versus the usual $O(n\sp2)$ (k is the number of input surfaces, n the number of finite elements into which the input surfaces are meshed). Similarly a hierarchical technique has been introduced for the more general radiance problem (which allows glossy reflectors) by Aupperle et al. (6). In this dissertation we show the equivalence of these hierarchical techniques to the use of a Haar wavelet basis in a general Galerkin framework. By so doing, we come to a deeper understanding of the properties of the numerical approximations used and are able to extend the hierarchical techniques to higher orders. In particular, we show the correspondence of the geometric arguments underlying hierarchical methods to the theory of Calderon-Zygmund operators and their sparse realization in wavelet bases. The resulting wavelet algorithms for radiosity and radiance are analyzed and numerical results achieved with our implementation are reported. We find that the resulting algorithms achieve smaller and smoother errors at equivalent work.

Illumination Invariant Elastic Bunch Graph Matching for Efficient Face Recognition

Abstract: This paper proposes a system which can handle illumination problem of face recognition systems by using "Retinex and color constancy" algorithm. The Retinex and color constancy approach has been plugged with Elastic Bunch Graph Matching (EBGM). The proposed system has been tested on IITK database having more than 1000 face images. The experimental results demonstrate that performance of the proposed system is superior to the known systems. The overall accuracy has shown an increase of 3.14% as compared to the known EBGM based recognition system without using Retinex and Color Constancy method.

Polygonal Billiards with Small Obstacles

Abstract: The technique of unfolding a polygonal billiard table is used to answer certain questions concerning the illumination problem The main problem addressed is how many point obstacles would suffice to block any billiard path between two points of the polygon The answer can then be generalized from point obstacles to small ∈-neighborhoods of points

Feature extraction based on fuzzy local discriminant embedding with applications to face recognition

Abstract: Recently, local discriminant embedding (LDE) was proposed to manifold learning and pattern classification. LDE achieves good discriminating performance by integrating the information of neighbour and class relations between data points. However, in the real-world applications, the performances of face recognition are always affected by variations in illumination conditions and different facial expressions. LDE still cannot solve illumination problem in face recognition. In this study, the fuzzy local discriminant embedding (FLDE) algorithm is proposed, in which the fuzzy k -nearest neighbour (FKNN) is implemented to reduce these outer effects to obtain the correct local distribution information to persuit good performance. In the proposed method, a membership degree matrix is firstly calculated using FKNN, then the membership degree is incorporated into the definition of the Laplacian scatter matrix to obtain the fuzzy Laplacian scatter matrix. The optimal projections of FLDE can be obtained by solving a generalised eigenfunction. Experimental results on ORL, Yale and AR face databases show the effectiveness of the proposed method.