Process-in-Network for Image Providing Services

Abstract

Most image processing techniques are based on the idea that the information is processed in end nodes. Our proposal is based on a slightly different idea called Process-in-Network (PIN) defined as the possibility that the network processes information as it is being transmitted, seeking to take advantage of waiting times, idle processing capacity, and the information itself. In this article, the possible use of PIN in image processing and analysis applications is discussed. A simple image processing example is first presented to illustrate PIN concept, and then other possible application cases are also discussed, exploring the possibility of PIN to adapt a picture or video to different QoS requirements, and a character recognition application for the use of a transit department. A feedback mechanism could be implemented in order to dynamically adapt QoS or QoE for an specific user at the destination node. The approach here presented may be generalized not only for image processing applications but also for video processing, in which case it could be complemented with the use of other technologies and standards such as SVC and MapReduce.