In the on-board image coding algorithm,one of the main problems in ROI (Region-of-interest) coding is to decide the scaling value. A novel algorithm for calculating entropy estimate is presented which can approximate to the actual output rate produced by bit-plane coding. Then an algorithm for deciding the scaling value of ROI is proposed by which the scaling value of ROI can be decided correctly. The scaling value is calculated according to the image being coded and the requirement,which does not need inference,with different image complexity different lifted value. So the algorithm in the paper is suitable for those images such as multi-spectral images in which the ROI position is known and their complexity different. The simulation results show that in spite of different complexity the algorithm can decide the lifted value precisely so that the whole reconstructed image quality is better.