Showing papers on "Illumination problem published in 2018"
31 Jan 2018
TL;DR: To solve the illumination problem of the conventional face recognition system using Haarcascade algorithm, LBPH is merged into the system with the HOG linear SVM object detector, in this paper.
Abstract: To solve the illumination problem of the conventional face recognition system using Haarcascade algorithm, LBPH is merged into the system with the HOG linear SVM object detector, in this paper.
4 citations
Patent•
21 Dec 2018
TL;DR: Zhang et al. as mentioned in this paper provided a construction method for non-negative feature extraction and face recognition application, which comprises the following steps: characterizing loss degree by cosine measure, characterising loss degree after matrix decomposition, determining loss degree between matrices, and obtaining an update iteration formula: the objective function is transformed to form the optimization problem, and the updated iterative formula of the algorithm is obtained by constructing auxiliary function.
Abstract: The invention provides a construction method for non-negative feature extraction and face recognition application, which comprises the following steps: characterizing loss degree by cosine measure; characterizing loss degree after matrix decomposition by cosine measure between matrices; and determining loss degree by cosine measure between matrices. A method for constructing an objective functioncomprises that step of characterizing a loss degree by a cosine measure to form an objective function; obtaining an update iteration formula: the objective function is transformed to form the optimization problem to be solved, and the updated iterative formula of the algorithm is obtained by constructing auxiliary function. The invention has the advantages that: 1. the illumination problem encountered in the face recognition process is solved; 2. the convergence of the algorithm proposed by the invention is not only proved in theory by using auxiliary function, but also verified in experiment,and our algorithm has higher convergence; 3. compared with the related algorithm in the face database with illumination influence, the result shows that the algorithm of the invention has certain superiority.
4 citations
01 Jul 2018
TL;DR: Experiments on the Yale B, the extended Yale B and the CMU PIE face databases show that the proposed method provides better results than some state-of-the-art methods, showing its effectiveness for illumination normalization.
Abstract: The illumination problem is one of the main bottlenecks in a practical face recognition system. Illumination preprocessing is an effective way to handle lighting variations for robust face recognition. In this paper, we present a novel illumination insensitive image, namely directional gradients integration image (DGII), for illumination insensitive face recognition. Unlike the existing model-based methods, the DGII is generated directly from the decomposed gradient components of a logarithmic image, without involving any training procedure. Based on the Lambertian reflectance model, we first calculate the horizontal and vertical gradients in the logarithmic domain to eliminate the illumination component. Secondly, to utilize the gradient orientation information, the two gradients are further decomposed into four components along four directions. Then, the four directional gradients are integrated to reconstruct an illumination insensitive image using anisotropic diffusion. Finally, the reconstructed image is fused with the gradient magnitude image through weighted summing. For performance evaluation, we simply use principal component analysis for feature extraction, Euclidean distance as similarity measure and nearest-neighbor classifier for face recognition. Experiments on the Yale B, the extended Yale B and the CMU PIE (The Carnegie Mellon University pose, illumination and expression database) face databases show that the proposed method provides better results than some state-of-the-art methods, showing its effectiveness for illumination normalization.
1 citations
21 Dec 2018
TL;DR: This judicial combination of representation and classification shows improved recognition accuracy on benchmark databases and advantage of the proposed approach is also shown by the performance analysis with single training image, which is necessary for some real time applications.
Abstract: An approach of class specific representation based learning for illumination tolerant face recognition is reported in this paper. Autoencoder based representation and class specific reconstruction along with phase correlation in frequency domain for classification is proposed. Autoencoder based representation is evaluated as very few number of training images are sufficient to handle the entire variation of test face subspace. Phase correlation is used at the classification stage to handle the illumination problem as intensity is the primary concern. This judicial combination of representation and classification shows improved recognition accuracy on benchmark databases. The performance of the proposed approach compared to another state-of-the-art technique on other representation based learning is established with extensive experimental. Advantage of the proposed approach is also shown by the performance analysis with single training image, which is necessary for some real time applications.