Low-overhead video compression combining partial discrete cosine transform and compressed sensing in WMSNs

摘要

Wireless multimedia sensor network (WMSN) is a special wireless sensor network (WSN) made up of several multimedia sensor nodes, specially designed to retrieve multimedia content such as video and audio streams, still images, and scalar sensor data from the environment. Due to strict inherent limitations in terms of processing power, storage and bandwidth, data processing is a challenge in such network. Further, energy is one of the scarcest resources in WSN, especially in WMSN and therefore, saving energy is of utmost importance. Data compression is one of the solutions of such a problem. This paper proposes an energy saving video compression technique for WMSN by judicious combination of partial discrete cosine transform and compressed sensing. This amalgamation exploits the benefits of both the techniques towards fulfilling the objective of saving energy along with achieving desired peak signal to noise ratio (PSNR). When the transform technique ensures low-overhead compression, compressed sensing guarantees the reconstruction of the same video with lesser amount of measurements. Performance of the scheme is measured both qualitatively and quantitatively. In qualitative analysis, overhead of the scheme is measured in terms of storage, computation, and communication overheads and the results are compared with a number of existing schemes including the base scheme. The results show considerable reduction of all such overheads thereby justifying the appropriateness of the proposed scheme for resource-constrained networks like WMSNs. In quantitative analysis, for both ideal and packet loss environment, the scheme is simulated in Cooja, the Contiki network simulator to make it readily implementable in real life mote e.g. MICAz. When compared with the existing state-of-the-art schemes, it performs better not only in terms of 34.31% energy saving but also in getting an acceptable PSNR of 35-37 dB and SSIM of 0.85-0.88 in ideal environment. In packet loss environment, these values are 32.9-35.5 dB and 0.81-0.85 respectively implying acceptable reconstruction even in packet loss environment. Further, it requires the least storage of 51.2 KB. The observation on simulation results is also justified by statistical analysis.