Illumination invariant face recognition using dual-tree complex wavelet transform in logarithm domain

Abstract

In this article, we develop a new algorithm for illumination invariant face recognition. We first transform the face images to the logarithm domain, which makes the dark regions brighter. We then use dual-tree complex wavelet transform to generate face images that are approximately invariant to illumination changes and use collaborative representation-based classifier to classify the unknown faces to one known class. We set the approximation sub-band and the highest two DTCWT coefficient sub-bands to zero values before the inverse DTCWT transform is performed. Experimental results demonstrate that our proposed method improves upon a few existing methods under both the noise-free and noisy environments for the Extended Yale Face Database B and the CMU-PIE face database.

Noise Robust Illumination Invariant Face Recognition Via Bivariate Wavelet Shrinkage in Logarithm Domain

“…The correct classification rate is defined as the percentage of faces that are recognized correctly. In this paper, we only implement our proposed method in this paper, LOG-DTCWT [16] and LOG whereas all other classification results are copied from Xie et al [14]. For subset 1 of the Yale-B face database B, our method yields classification rate of 94.67% whereas Large and Small-Scale features [14], LTV [11] and Local Binary Pattern [10] approaches yield perfect classification rate (100%).…”

“…Figure 6 shows the noise-added face images for σ n ranging from 5 to 40. We compare our proposed algorithm with LOG-DTCWT [16], LOG-Discrete Wavelet Transform (LOG-DWT) and LOG-DCT for both extended Yale face database B and CMU-PIE face database. Table 3 shows the correct classification rates (%) of the proposed method, LOG-DTCWT [16], LOG-DWT, and LOG-DCT for face images corrupted by Gaussian white noise.…”

“…Xie et al [14] achieved good results by normalizing face illumination and used large-and small-scale features. Chen et al [15,16,17] investigated several novel algorithms based on dual-tree complex wavelet transform (DTCWT) and other filters. Chen et al [18,19] studied hyperspectral approaches taking advantage of log-polar Fourier and other features.…”

Handwritten Character Recognition Using Unsupervised Feature Selection and Multi Support Vector Machine Classifier

“…The splitting of subset is shown in Table 1. If the angle formed between light source and camera (azimuth and elevation) is less than 12 degrees, than it is categorized as Subset 1, Subset 2 are from 12 degrees through 25 degrees, Subset 3 are form 26 degrees until 50 degrees, Subset 4 are form 51 degrees to 77 degrees and Subset 5 are above 78 degrees [24]. Therefore, the Subset 5 contains face image with the most extreme variation in lighting conditions, while the Subset 1 contains the less lighting variation images.…”

Face Recognition State-of-the-art, Enablers, Challenges and Solutions: A Review

“…Shi et al [31] presented a collaborative-representation-based classification-algorithm (CRC) approach using statistical histogram measurement (H-CRC) in conjunction with a 3D morphable model (3DMM) for pose-invariant face classification. Chen et al [32] proposed an algorithm for addressing the illumination invariant face recognition. Dual-tree complex wavelet transform was used to generate face images that are approximately invariant to illumination changes.…”