- 期刊
Application of Novel Gabor-DCNN into RGB-D Face Recognition
摘要
In this paper, we propose a novel approach, named Gabor- DCNN, applied for face recognition technology of two modalities RGB-D images, which can extract the features through Gabor transform of images and deep convolutional neural network. Gabor transform can capture the salient visual properties with spatial frequencies and local structural features of different directions in a local area of images and enhance the object representation capability. Deep convolutional neural networks could automatically learn essential features from images compared with traditional methods. The most significant features can be obtained through the Gabor-DCNN. The final features expression of the face is performed by feature fusion of two modalities images. The experimental results indicate that the algorithm achieves much better performance than some state-of-the-art methods in terms of recognition rates on the EURECOM data set. This research provides an effective method for multiple modal face recognition under complex conditions.