Article Preview
Top1. Introduction
Texture analysis is important in many applications of computer image analysis and has been widely studied in the literature. Many existing approaches deal with limited problems that do not fully take into account the image variations with respect to orientation, intensity and spatial scale.
Over the years, description of texture has interested many researchers. Few have incorporated at least one property of invariance. Some of the work related to gray scale and rotation invariant classification of texture images is listed as follows. Chen and Kundu (1994) were among the first researchers to propose a rotation and gray scale transform invariant texture recognition scheme using the combination of quadrature mirror filter bank and hidden Markov model (HMM). This was followed by Wu and Wei (1996) who developed a method using spiral resampling, subband decomposition and hidden Markov model. Ojala, Pietikainen, and Maenpaa (2000) implemented a gray scale and rotation invariant texture classification method based on Local Binary Patterns (LBP) and non-parametric distribution of sample and prototype distributions. Ojala, Valkealahti, Oja, and Pietikainen (2001) used signed gray level differences and their multidimensional distributions for texture description. Dalal and Triggs (2005) evaluated Histograms of Oriented Gradients (HOG) for human detection. Pok, Pyu, and Lyu (2005) improved the spatial distribution of features by using correlations among local texture patterns for rotation and gray scale invariant texture classification. Zeng, Naghedolfeizi, Arora, Yousif, Gosukonda, and Aberra (2013) transformed a gray scale image into a pattern map using Gabor filters and then used it for pattern matching. Shahidul Islam and Auwatanamongkol (2013) derived a four-bit pattern based on gradient directions of the gray scale values of its neighbouring pixels for texture representation. Kou, Cheng, Chen and Zhao (2018) introduced a texture descriptor based on the principal curvatures and rotation invariant version of completed local binary pattern. Liu, Liu and Chen (2019) blended pairwise rotation invariant co-occurrence local binary pattern with local texture information for rotation and illumination invariant texture description.
The texture analysis community has developed a variety of different descriptors for the appearance of facial image patches. The objective of facial feature representation is to derive a set of features from original face images that is invariant to changes of rotation and gray scale / illumination. Ahonen, Hadid, and Pietikainen (2006) applied Local Binary Patterns for face recognition with outstanding results for rotation invariance. Zhang, Tang, Fang, Shang, and Liu (2009) derived Gradientfaces from the image gradient domain for face recognition under varying illumination. Deniz, Bueno, Salido, and Torre (2011) used Histograms of Oriented Gradients (HOG) for face recognition. Huang, Shan, Ardebilian, Wang, and Chen (2011) surveyed the use of Local Binary Patterns and its variants for facial image analysis. Huang and Yin (2017) computed image gradients from multiple directions and simplified them into a set of binary strings for face recognition. Zhou, Constantinides, Huang, and Zhang (2017) fused together center symmetric Local Binary Pattern and image entropy for illumination invariant face recognition. Alotaibi, Alharbi, and Kurdi (2017) used homomorphic filter and Local Binary Patterns for face recognition in illumination variations.