According to the characteristics of the remote sensing images for ocean surveillance satellite,an adaptive near-lossless compression algorithm based on ROI(Regions of Interest) is proposed.After maximum lifting of morphological Harr wavelet,the objects are detected by threshold method and connectivity analysis of eight adjacent regions.The ROI of high-frequency sub-bands is depicted by the intersection of enclosing rectangle and an annulus in the lowest-resolution and gained by mosaic magnification in other resolution levels.In high-frequency sub-bands,Rice lossless entropy encoder is used for ROI,and bit plane encoder for background region,while DPCM and Rice for low-frequency sub-band.The experimental results show that the algorithm can segmentalize the ROI effectively without any visible segmentation trace.The PSNR(Peak Signal Noise Ratio) of proposed algorithm is 2～5 dB higher than that of JPEG2000 at low and middle bit rate,and both are higher than 45 dB at high bit rate.The algorithm has the advantages of low complexity,higher adaptive ability and independent datum packet and is easy to be implemented in hardware.Moreover,the algorithm can adapt to near-lossless compression for ocean surveillance satellite image.