Texture Feature Extraction of Flotation Foam Image Based on Gabor Wavelet and Improved LBP

Flotation bubble images under different conditions but with similar texture structures pose recognition challenges. Aiming to address this, we propose a texture feature extraction approach for flotation bubble images based on Gabor wavelets and an improved local binary pattern (LBP) method. Unlike traditional LBP which overlooks center pixel features and regional amplitude traits, the improved LBP incorporates new LBP_C and LBP_M operators to account for this extra information. Gabor wavelets are leveraged to perform multi-scale and multi-directional filter decomposition on bubble images, followed by computing the Gabor amplitude feature GM for each decomposition. LBP_S, LBP_C and LBPJM operators are then utilized to calculate feature spectra of each GM, which are concatenated to obtain texture feature vectors for the bubble images. Finally, a support vector machine (SVM) is trained to categorize bubble images under different conditions. Results validate that the proposed approach significantly improves classification accuracy between various flotation bubble states, demonstrating its superiority in texture feature extraction.