Determining the Region of Single-Target Interest Area by Prediction Method in Wireless Multimedia Sensor Networks

摘要

Gray values have always been altered in images collected by wireless multimedia sensor networks because of changes in light, weather and other conditions of monitored environment. In this case it may lead to the non-interest areas in the images to be misjudged as interest areas. If there are only a small number of pixels that have been affected in an image, the probability of misjudgment is smaller. However, if the affected pixels are massive, the difference method may judge many of non-interest areas as interest areas by mistake. This will increase the energy consumption of image compression process. Besides, it would not help to improve image quality. Therefore in the case of fixed reference frame, when there is an abrupt change in background environment, and only one concerning target in the image, we propose a method to predict the interest areas of current frame image by using the interest areas of history frames and the movement trend of the moving target. Binary-conversion based on the interest area and background area on the previous two historical frames. By using the connected component labeling algorithm based on run-length coding a single-target interest area can be determined. Predictor is determined according to coordinates' extremes of two frames, and then the interest area of the current frame is predicted according to the previous frame and the predictor.