Local Binary Pattern(LBP) Optimization for Feature Extraction

TL;DR

This work reframes Local Binary Pattern (LBP) feature extraction as a triple-matrix operation $F = T\,E\,H$ and leverages singular value decomposition (SVD) to learn problem-specific transformation matrices $T$ and $H$ that maximize class separability between mean LBP matrices $\bar{E}_1$ and $\bar{E}_2$. It also introduces an algorithm to identify and refine optimal LBP codes, producing features that are highly discriminative while using far fewer features than standard LBP. Experimental results on face detection and facial expression recognition (across CFD-T, UTKFace, and CK-Data) show substantial gains in accuracy with compact feature vectors, validating the proposed approach. The method offers a principled path to dataset-tailored texture descriptors and can be extended to larger neighborhoods and other descriptors.

Abstract

The rapid growth of image data has led to the development of advanced image processing and computer vision techniques, which are crucial in various applications such as image classification, image segmentation, and pattern recognition. Texture is an important feature that has been widely used in many image processing tasks. Therefore, analyzing and understanding texture plays a pivotal role in image analysis and understanding.Local binary pattern (LBP) is a powerful operator that describes the local texture features of images. This paper provides a novel mathematical representation of the LBP by separating the operator into three matrices, two of which are always fixed and do not depend on the input data. These fixed matrices are analyzed in depth, and a new algorithm is proposed to optimize them for improved classification performance. The optimization process is based on the singular value decomposition (SVD) algorithm. As a result, the authors present optimal LBPs that effectively describe the texture of human face images. Several experiment results presented in this paper convincingly verify the efficiency and superiority of the optimized LBPs for face detection and facial expression recognition tasks.

Figures (11)

• Figure 1: LBP values computation process. Each $3\times 3$ pixel block in the image is encoded to a LBP value.
• Figure 2: LBP feature extraction process. The image is divided into 16 sub-regions and LBP values are extracted from each region. Then the histogram is applied on the LBP values to extract $8$ features for each sub-region. Then, finally, $128=16 \times 8$ features are exploited from the image as LBP features.
• Figure 3: Function $lbp^{-1}$ transforms $lbp$ values to 8-bit binary numbers. One can see $i$ end cells of the numbers are only important when the histogram divides the $lbp$ values into $2^i$ equal sub-intervals for $i=1, 2, ...,8$.
• Figure 4: The extension vector for a LBP value. The vector takes 1 at $lbp$-th cell and 0, otherwise.
• Figure 5: The first vector of the transform matrices, $T$ and $H$, for the standard LBP, SVD, and the proposed LBP are presented from left to right. The first and second rows are respect to the values of $H$ and $T$ matrices, respectively.

Theorems & Definitions (6)

• Theorem 5.1
• proof
• Corollary 5.1
• Theorem 5.2
• proof
• Corollary 5.2