Abstract: Many methods only consider global but not key local features in age estimation tasks. A facial age estimation method based on double-enhanced features is proposed. Facial images are partitioned according to the location of the key points. Three local areas including eye, nose and mouth containing age related features (such as eye wrinkles, nasolabial folds, moustache and so on) are obtained. These local regions can enhance key local feature information on the basis of global features. The local area and the whole image are combined as the input of the parallel residual network based on the squeeze and excitation. The network further enhances the useful features and suppress those non-related ones by feature re-calibration. The estimated results of multiple sub-regions are combined as the final output. Experimental results indicate that the age estimation using global and local features outperforms those use global features only. The proposed method provides lower mean absolute error and computational complexity compared with others.