Texture is always considered as the preconscious for human vision. Texture also remains the same significance in computer vision field that can be used to help in detection, segmentation and classification tasks. Since texture is a global feature inherent in an image, containing essential surface information, which can be described in detail and hardly affected by image noises. We propose a novel end-to-end structure to make use of hybrid features by a mixture network and improve the classification accuracy, mainly combining Gray Level Co-occurrence Matrix (GLCM) statistical features together with pyramid structured deep convolutional neural networks (Pyramid CNNs) features in a paralleling network structure. Considering GLCM is a remarkable descriptor for texture statistical features, it can compensate the missing information in the convolution and pooling process of CNN and decline overfitting problems. Meanwhile, multi-resolution image pyramid structured CNN helps to capture both global features and local features. Quantitively, we carry out experiments on widely used datasets and results show that the GLCM and Pyramid CNN features merged structure obtains maximum 6.8% improvement comparing to the basic CNN methods.