This paper presents a new low computational power compression/decompression algorithm for very low bit-rate video-stream integrated on a standard architecture for domotic sensor network based on the IEEE 1451 standards. The proposed method uses frame theory elements and the greedy Matching Pursuit algorithm in order to identify a 2D Gabor sub-dictionary able to compress information included in little 8bpp grey-levels motion blocks obtained by displaced frame difference approach. Results show high performance in terms of compression ratio, good detected visual quality in the decompressed video stream, providing high signal to noise ratio.