3-D Compression-Oriented Image Content Correlation Model for Wireless Visual Sensor Networks

Wireless visual sensor networks (WVSNs) comprise a large number of camera-equipped visual sensors, which are often deployed densely to gain enhanced observations in field of interest. Therefore, there exists abundant image redundancy caused by the image content correlation (ICC) between cameras with the overlapped field of views (FoVs). Image compression is an important method to remove the image redundancy. Aiming at the problem of image compression, this paper designs a 3-D compression-oriented ICC (3D-COICC) model for WVSNs to quantitatively describe the ICC characteristics. First, the 3-D sensing model is developed by providing an improved method for the calculation of FoVs. After that, the 3D-COICC model is proposed by comparing the difference between foreshortening effects for one observed object, and then an algorithm of 3D-ICC is provided to estimate the ICC value. Furthermore, a scheme of correlation-based compression (CBC) is proposed to remove the redundancy of content between transmitted images, and thus the communication data are compressed. Where, an algorithm of image block restoration is provided to realize the transformation between two image blocks with the same pre-image. Evaluation results demonstrate that the developed 3-D sensing model can accurately depict the FoV of camera, and the 3D-COICC model outperforms the state of the art in accuracy. Further simulations show that the CBC scheme based on the 3D-COICC model is an effective method to remove the image content redundancy and improve the compression ratio.