This paper considers the problems in generating a single image from a combination of multiple images using a group of mobile devices with limited computational resources. The method of forming composite images using a group of mobile devices with limited computational resources is formulated. In particular, the model of organization of distributed computational process in a group of mobile devices is proposed and considered. The question of providing flexibility and stability of the system for the possibility of creating different groups of devices for different configuration of tasks and for situations of failure of one or more devices is considered. The problems of optimal information gathering by a group of mobile devices and optimal distribution of homogeneous tasks were also investigated. It was proposed to use Key.Net neural network to preserve the advantages of classical algorithms for determining key points while reducing time costs. The Key.Net neural network model was optimized using quantization to reduce the use of device memory and reduce the computational complexity of the proposed solution. To confirm the correctness of the obtained results, experiments were conducted, including the comparison of the initial and obtained models by various criteria. The results of the experiments showed the effectiveness of the applied approach.