The hybrid coding of motion compensated frame difference and transform is the basic structure of the image coder for video transmission at sub-T1 rate transmission rate.
By reducing the variance of the MCFDS(Motion Compensated Frame Difference Signals) in this coder, then number of image blocks to be transmitted can be reduced.
In this thesis a method of blockwise frame difference estimation - the MCFDS of current block is estimated with the neighboring MCFDS which was already transmitted - is proposed to reduce the variance of MCFDS effectly.
Two approaches are applied for estimation gain. The one is a linear estimation based on the covariance of MCFDS. It is a function of motion estimation error, which is also related to the mean of temporal covariance of original image. The second method used is a recursive estimation method.
When the proposed method is applied to the motion compensated hybrid coder using conditional replenishment technique, it is found that NMSE(Normalized Mean Square Error) gain of 1.5dB is achieved and bit rate can be also reduced above 18%.
