With the increasing popularity of digital media along with the fact that massive image data has already gained popularity, the automatic image recognition and classification have become a very critical and urgent problem, this study investigates and tests intelligent image recognition and classification techniques for digital media based on major methods such as CNN, SVM and RNN. Different algorithms and models are applied to training and test image data to evaluate the performance of various methods in terms of accuracy and processing time. The experimental results shows that the CNN method can achieve high accuracy in image classification task and the multi-layer convolution and pooling operations can effectively extract image features. The SVM method is better than the CNN in terms of processing time but have slightly lower accuracy. The RNN method is ranked between the CNN and SVM in terms of accuracy and processing time.