Object-oriented image sequence coding using the minimum description length principle

It is evident that there are increasing demands for ultra-low bit-rate video image transmission. We present a new formulation of moving object segmentation and estimation for the purpose of object-oriented coding schemes. Motivated by the minimum description length (MDL) principle´s unified treatment of data modelling and parameter estimation, a global optimal criterion is constructed in terms of the ideal coding length for each frame image. An optimization procedure to segment and estimate moving objects in a scene is proposed using a region-merging technique. Each object in the scene is represented by boundaries, motion parameters, and the motion-compensated prediction error. A number of experimental comparisons between block-oriented and object-oriented coding schemes are made to assess the coding gain using the new MDL-based method