This report discusses how to detect water level of a
river from video signal for the purpose of environmental
surveillance. Video signal is assumed to be taken 24 hours a day
with a web camera installed to the river side. A DSP inside the
video sensor is supposed to send the water level data regularly
and video signal irregularly on demand. It is our purpose to
utilize JPEG 2000 (JP2K) core technology for not only
"compression" of video data but also "recognition" of river's
water level so that redundancy of signal processing of the video
sensor can be reduced.