Research On Face Recognition Algorithm Based On Local Patterns

With the rapid development of biometric identification technology, many identification technologies emerge, bringing great convenience to people. It includes iris recognition, fingerprint recognition, gesture recognition, face recognition and so on. Among all the technologies, face recognition evolves into a mainstream recognition technology because of its non-contact, stability, simple operation etc.The key aspects of the face recognition are feature extraction and classification. The result of feature extraction determines the classification and it is a key step in face recognition. Therefore, this article studied the feature extraction in detail, analysis many recognition algorithm and improved face recognition based on local patterns.There are many algorithms in face recognition based on local patterns. The commonly used algorithms are based on local binary pattern algorithm, its abbreviation is LBP. Due to its simple calculation, it has been used widely. There are a lot of deformations such as local ternary mode (Local Ternary Pattern). In this paper, we improved the shortcoming of LTP algorithm and proposed a LTP algorithm based on dynamic threshold. For the reducing of the dimension of the feature vector extracted, we integrated two classic algorithms. Experiments show that, the improved algorithm is better than the original one.Gabor transform algorithm is a very traditional algorithm based on local patterns. There are many papers emerging on its deformation algorithm, such as Log-Gabor. This paper proposed a new algorithm based on Log-Gabor energy value algorithm to solve the shortage of the paper which used statistical sampling on Log-Gabor. We analyzed the algorithm from the local, orientation, scale three angles and designed three filter banks. Experiments show that, the improved algorithm increased the recognition rate.