Comparative Study between Texture Feature and Local Feature Descriptors for Silk Fabric Pattern Image Recognition
Thananchai Khamket and Olarik Surinta
Abstract
Thai silk fabrics have unique patterns in different regions of Thailand. The designers may have been inspired and took ideas from the natural environment to create new silk patterns. Hence, many new silk patterns are modified from the original silk pattern. It is challenging for people to recognize a pattern without any prior knowledge and expertise. This paper aims to present a comparative study between texture feature and local feature descriptor for silk pattern image recognition. First, two feature extraction techniques: texture feature and local feature descriptors are proposed to create robustness features from sub-regions that are divided by the grid-based method. Second, the robust features are then classified using the well-known and effective classifier algorithms: K-nearest neighbor (KNN) and support vector machine (SVM) with the radial basis function. We experimented with silk pattern image recognition on two silk fabric pattern image datasets: the Silk-Pattern and Silk-Diff-Pattern. The evaluation results show that the texture feature called the local binary pattern (LBP) when combined with the KNN and SVM algorithms outperforms other feature extraction methods, even deep learning architectures.
Keywords — Texture feature; Local feature descriptor; Silk fabric pattern image recognition; Support vector making; K-nearest neighbor.
Read article — https://dl.acm.org/doi/abs/10.1145/3388176.3388201
Conclusion
In this paper, we compared methods of silk pattern image recognition. First, we proposed a grid-based method for dividing the image into sub-areas. Second, the two well-known feature extraction techniques: texture feature and local feature descriptor, were proposed to extract robust features from each sub-area. The grid-based method allowed the feature extraction technique to extract more useful features. Finally, we concatenated the robust features and fed them to the classifier algorithms: K-nearest neighbor and the support vector machine algorithms.
The results showed that the LBP algorithm outperforms other methods when combined with both the KNN and the SVM algorithms. On the Silk-Pattern dataset, the LBP+KNN method performs much better than the other methods for all test sets. Subsequently, the LBP+SVM method shows the best performance on the Silk-Diff-Pattern dataset. We also compared our results with two basic deep learning architectures: LeNet and AlexNet architectures. We found that the deep learning architectures showed low accuracies when the training set is inadequate. To the best of our knowledge, the scale-invariant feature transform (SIFT) algorithm always showed the best performance. On the other hand, surprisingly, the SIFT algorithm had very low performance when combined with KNN and SVM algorithms on the silk image dataset.
We conclude that deep learning architecture obtain high accuracy on many pattern recognition problems. In future work, we want to study deep learning architecture and apply it to the silk image dataset. We are also interested to improve the performance by using transfer learning and data augmentation.