Preprocessing techniques based on LBP and Gabor filters for clothing classification

This paper presents preprocessing techniques for clothing classification system which are based on local binary pattern (LBP) and Gabor filters. Clothing are non-rigid and deformable objects that are very difficult for classification even human. Finding the appropriate features are one of the important issues for classification and remain challenges. LBP-based and Gabor-based methods are adopting to preprocessing for generating the beneficial information, including to analyze the properties of clothing. There are four preprocessing output images that generated before feature extraction. These images are LBP-based image, maximum magnitude image and combined information between maximum magnitude and gray scale image by division and subtraction operators. In the experiments, we extracted and analyzed feature properties by comparing between preprocessing and non-preprocessing images. Entropy, uniformity and LBP are applied to feature extraction system. By using the preprocessing techniques, the appropriate features have been extracted. The experimental results show that the preprocessing techniques can improve the performance of classification.