Progressive DCT Image Compression Based On Quad Tree Decomposition for WSN

Wireless Sensor Networks (WSNs) have nodes with limited energy, limited processing power, and limited memory. In addition, they have low bandwidth and noisy communication channels. Therefore, image compression is very important for WSNs to reduce the time or the required bandwidth for transmitting the images. Progressive image compression is useful in WSNs because even when there is a cutoff in the transmitted data due to ending the battery lifetime or channel noise, the decoder can decompress the partially received data to obtain an acceptable image. Unfortunately, the progressive Discrete Cosine Transform (DCT) standard is unsuitable for WSNs because it uses the arithmetic coder or the Huffman coding that is very complex to implement. In this paper, we replace these coders with the Quad-Tree Decomposition (QTD) which has very low complexity. Unfortunately, most of the algorithms that use the QTD require lists to record the state of the image pixels and the quad-trees. These lists require an additional memory of about (2-3) of the image size. Therefore, it is not suitable for WSNs. In this research, we present the No List-QTD (NL-QTD) image compression algorithm that solves this memory constraint. It employs marker bits of an average memory of about 1/10 the image size instead of the lists to mark these states. The low complexity of the proposed NL-QTD also preserves the energy of the sensor's battery which in turn increases its lifetime. The simulation results show that the proposed method also has a higher Peak Signal-to-Noise Ratio (PSNR), and a higher Structural Similarity Index Measurement (SSIM), as compared with other related algorithms.