While providing massive information, the intelligent media Internet of Things (IoT) also poses challenges to the overall environment and the development of modern market economy. The employment of enterprises and people is still facing great difficulties, and the world economic situation is still complicated and severe. In addition, there are many design resources for icon art design on the Internet, and with different design styles, the demand for icon design is also increasing. The biggest difference between icons and ordinary pictures is that icons can convey the characteristics and meaning of pictures faster. The Generative Adversarial Network (GAN) technology in intelligent image processing and the TensorFlow learning framework are used to build and improve the icon generation network to simplify the icon design process. Computers are used in place of designers for icon art design. Firstly, the related technical background of icon generation network implementation is drawn through the introduction of related concepts of intelligent image processing. Secondly, Python is used to process the established icon dataset. Finally, the icon generation network is improved. The model training results show that the icon generation network has a peak feature loss value of 9.0 and an average error of 8.0. After the color label is added, the effect is significantly improved. The improved icon generation network has a peak feature loss of 7.0 and an average error of 6.0. The results show that after the color labels are added, the improved GAN model has a very high recognition rate for artistic icons. The improved network model also distinguishes the newly generated icons from the original ones. The comprehensive application effect of the model is good. This provides specific application and reference value for the intelligent development of the IoT.

In terms of image processing, encryption plays the main role in the field of image transmission. Using one algorithm of deep learning (DL), such as neural network backpropagation, increases the performance of encryption by learning the parameters and weights derived from the image itself. The use of more than one layer in the neural network improves the performance of the algorithm. Also, in the process of image encryption, randomness is an important component, especially when used by smart learning methods. Deep neural networks are related to pixels used to manipulate position and value according to the predicted new value given from a variable neural system. It also includes messy encrypted images used via applying randomness and increasing the key space in addition to using the logistic and Henon map for complexity. The main goal of any encryption method is to increase the complexity of the encrypted image to be difficult or impossible to decrypt the image without the proposed key. One of the important measurements for image encryption is the histogram and how it can be uniformed by the proposed method. Variables of randomness are used as features for the deep learning system, with feedback during iteration. An ideal image processing encryption yields high messy images by keeping the quality. Experimental results showed the backpropagation algorithm achieved better results than other algorithms.