Abstract:Local Binary Pattern (LBP) is widely used in texture classification because of its powerful capability to extract texture features of a center pixel. However, LBP has three main drawbacks: (1) limited by the low resolution of imaging device, the quality of texture image is degraded, and some real existing pixels with more texture information are unavoidably lost. (2) Center pixel <math>gc</math> is the most important factor to extract correct LBP pattern. However, by far LBP and its variants do not show any solutions to enhance the robustness of center pixel <math>gc</math>. (3) Some LBP patterns include important texture microstructures, but they are ignored by uniform patterns. At the same time, some uniform patterns can be corrupted by noise and misclassified into non-uniform patterns. These LBP patterns therefore all lost their discrimination capability. In order to overcome these three disadvantages, in this paper, we propose a novel adaptive center pixel selection (ACPS) strategy.Inspired by image super-resolution techniques, ACPS firstly applies the interpolation method to recover the lost real existing pixels and generate the center pixel candidates with more texture information. Then, the gradient information is used to obtain the edge image aiming to find the non-uniform patterns at edge points which may contain complicated texture microstructures. After generating the center pixel candidates and edge image, we introduce ACPS strategy into the LBP framework. By adaptively selecting the optimal center pixel from all center pixel candidates, one non-uniform pattern at the edge point can be possibly recovered to the uniform pattern, and regain its discrimination power. It is worth noting that any other LBP variants can also employ the ACPS strategy to more effectively extract its texture features. By observing the experimental results on representative texture databases of Outex, UIUC, CUReT, XU_HR, ALOT and KTHTIPS2b after introducing the ACPS strategy into LBP and its variants of LTP, CLBP, BRINT, CRDP, FbLBP, and CJLBP, the texture classification performances can be significantly improved.

Distinguishing Structures from Textures by Patch‐based Contrasts around Pixels for High‐quality and Efficient Texture filtering

Edge-aware texture filtering with superpixels constraint

Improved Bilateral Texture Filtering With Edge-Aware Measurement

Patch-Based Approach for the Fusion of Low-Light Image Pairs

A Shape-Aware Structure-Preserving Texture Smoothing Algorithm

Multi-scale selective image texture smoothing via intuitive single clicks

Pointwise Shape-Adaptive Texture Filtering

Constant Time Texture Filtering

Structure-Texture Image Decomposition Using Discriminative Patch Recurrence.

Patch-based co-occurrence filter with fast adaptive kernel

Texture Synthesis Based on Feature Description

PatchScaler: An Efficient Patch-Independent Diffusion Model for Image Super-Resolution

Structured textons for texture classification

EDGE-PRESERVING IMAGE SMOOTH ALGORITHM BASED ON LOCAL EXTREMA INTERPOLATION

Adaptive center pixel selection strategy in Local Binary Pattern for texture classification

CONTRAST ENHANCEMENT USING TEXTURE HISTOGRAM AND FUZZY ENTROPY

An Edge-Located Uniform Pattern Recovery Mechanism Using Statistical Feature-Based Optimal Center Pixel Selection Strategy for Local Binary Pattern.

Footprint Area Sampled Texturing.

Contrast Enhancement Via Dual Graph Total Variation-Based Image Decomposition.

Generating High-Fidelity Texture in RGB-D Reconstruction Using Patches Density Regularization

Robust Contrast Enhancement Via Graph-Based Cartoon-Texture Decomposition