Moving Object Detection and Compression in Infra-Red (IR) Sequences

We consider the problem of remote surveillance using infrared (IR) sensors. The aim is to use IR image sequences to detect moving objects (humans or vehicles), and to transmit a few “best view images” of every new object that is detected. Since the available bandwidth is usually low, if the object chip is big, it needs to be compressed before being transmitted. Due to low computational power of computing devices attached to the sensor, the algorithms should be computationally simple.

We also present techniques for selecting a single best view object chip and computationally simple techniques for compressing it to very low bit rates due to the channel bandwidth constraint. A fast image chip compression scheme was implemented in the wavelet domain by combining a non-iterative zerotree coding method with 2D-DPCM for both low and high frequency subbands. Comparisons with some existing schemes are also included.

Papers