Performance Analysis of Bi-Level Image Compression Methods for Machine Vision Embedded Applications

Wireless Visual Sensor Network (WVSN) is an emerging field which combines an image sensor, an embedded computational unit, a wireless communication link, memory and energy resource. The individual Visual Sensor Nodes (VSNs) acquire image of the area of interest, perform some local processing on it and transmit the results using an embedded wireless transceiver. The processing of images at the VSN requires higher computing power and their transmission requires large bandwidth. Normally, the WVSNs are deployed in remote areas where the installation of wiring for power and data transmission is either not feasible or expensive. Due to the unavailability of continuous power, the energy budget in WVSN is limited. The limited energy budget requires that the processing at the VSNs and the communication to the server should consume as little energy as possible. The transmission of raw images wirelessly consumes a great deal of energy. Data compression methods can efficiently reduce the data and will thus be effective in reducing the communication energy consumption of the VSN. This paper explores seven well known bi-level image compression methods based on their processing complexity on the embedded platforms. The focus is to determine a compression method which can efficiently compress bi-level images and its processing complexity is suitable for the embedded platforms usually used for the implementation of the VSN. This paper is intended to be a resource for the researcher interested in using bi-level image compression methods in energy constrained real time embedded systems.