Rate Control Method for Video Encoders Operating in Industrial Environments

Video coding technology aims to reduce the bandwidth needed for video transmission and storage while maintaining the maximum possible video content quality. Rate Control (RC) is the mechanism designed for keeping the compressed rate at some predefined level. While in generic video coding, this has been thoroughly investigated in the past, in industrial computer vision applications using fixed cameras, there are not many efficient solutions capable of reaching the target rate in a fast and efficient manner. This paper proposes a RC method specifically tailored for industrial applications where the variability of the visual content is low and transmission to central servers is prone to errors and losses. The proposed algorithm is adaptive to the scene to be encoded through a startup procedure that builds default rate-quantization functions. Then, in normal operation, these functions are used to determine the quantization parameter that produces the best matching rate considering the predefined target, which can be dynamically defined from network or storage constraints. The results show that steady-state operation is achieved with low absolute error (1.08%) with respect to the target bitrate (BR) and also with fast setting time when abrupt changes are imposed. Overall the proposed method is flexible enough to match the requirements of a wide variety of industrial applications, namely those where remote video should be delivered in compressed format for cloud processing or storage servers.