The Offset-Image Recognition Based on Dense Coding

Junxiao Cao, Ziyu Yao, Jinjin Chen, Yuanyuan Zhang, Xingze Tai, Tian Zhang, Jiaquan Gao

Abstract: In this paper, the influence of dense coding on the recognition of offset-image is explored by improving the feature extraction ability of network model. By the experimental results, we find that adding the PCANet convolution layer to obtain more filters cannot effectively improve the network recognition ability of the offset-image recognition. Based on this background, we propose a new dense network framework on the basis of the FPH framework. Experimental results on the AR dataset and MNIST variations show that the proposed network framework is effective and improves significantly the ability of the offset-image recognition.