Certain investigation on transform based compression schemes for wireless image communication

Abstract

Wireless Sensor Networks (WSN) are composed of battery-powered video cameras, each packaged with a low-power wireless transmitter and receiver which is capable of processing, sending, and receiving data. Video and audio sensors are used to enhance the surveillance systems. Transmission of images and videos in wireless sensor networks consists of problems like lack of priori, fading, unpredictable signal strength, dynamically changing bandwidth and limited life time of the battery used in the transmission nodes. In order to deal with the above issues, the most important factor to be addressed is the energy of the battery, because most of the energy is used for communication between the nodes. Most of the battery power is used in video and image transmission. So, it is important to have an efficient compression algorithm. The compressed image thus can combat the poor transmission nodes. But this transmission comes with error in the retrieved image (decoded image) or often the image s important features are compressed to such an extent that it cannot be used. This is due to the fact that any image compression method degrades the important artifacts /features of the image. Numerous image compression algorithms are proposed for wireless sensor networks, but the biggest challenge is the lifetime of the battery in the network. This is because most of the WSN works in remote area and it is very difficult to replace the battery. Considering wireless sensor networks operates at resource constraint environments many image communication algorithms were proposed, but there is always a chance of optimization in the energy consumed by the network for image transmission. Ten years from now onwards this thesis will not be read like this, instead will watch it as a video because picture gives us more information. In the coming years, the world s internet traffic will be mostly videos. But transmitting images and videos requires more space and computations which creates burden to the wireless image communication. So, compression plays a vital role in image communication. Lossy image compression compresses the whole image with similar filter coefficients causing extreme degradation of the image quality. So, when the image is reconstructed lots of artifacts are found in the image. To reduce this problem the image is compressed based on the Region of Interest (ROI) and keep the ROI minimally changed. The first proposed method in this thesis is Shape Adaptive Discrete Cosine Transform (SADCT).This method uses a predefined orthogonal sets of DCT s whose parameters are stiff (Compression Ratio is higher) for regions outside the ROI and the parameters are laxed (Compression ratio is lesser) for the portions of image inside the ROI. Hence the outcome of the compression method gives lesser error inside the ROI. ROI is obtained from the saliency maps of colour images. newline