Comparison of encoding techniques for transmission of image data obtained using compressed sensing in wireless sensor networks

The Wireless Sensor Network (WSN) has limitations in bandwidth and computational resources as they have limited communication and storage capabilities. WSN consists of cameras, which have some local image processing and one or more central computers, where image data from multiple cameras is further processed and fused. Because of these limitations, the encoding techniques used for transmitting the image data should be efficient in order to make use of the available resources properly. A new sampling method is also introduced in the Image/video encoder of the WSN called Compressed Sensing (CS), which is the process of acquiring and reconstructing a signal that is supposed to be sparse or compressible, thus reducing the computational complexity. The image is divided into dense and sparse components by applying 2 levels of wavelet transform. The dense component uses the standard encoding procedure such as JPEG and the sparse measurements obtained from the sparse components are encoded by the techniques such as Exponential Golomb coding followed by Run-length encoding and arithmetic coding and the performances in terms of compression ratio and bits per pixel are compared. The recovery algorithm may be anyone supporting the compressed sensing technique such as OMP, POCS etc. In this work, the measurements (used in CS) and the predicted sparse components as the initial values, the projection onto convex set (POCS) recovery algorithm is used to get back the original sparse components of two levels and hence the original image by applying the inverse of transform to the dense and recovered sparse components.