Aiming at the application of the wireless image transmission for target de tection and identification,in order to solve the conflict between the requirement for high image quality in detec tion and recognition task and the constraint of wireless channel bandwidth,a region of interest (ROI) progressive coding algorithm based on multiscale compressed sensing is proposed.First,the technology of multiscale compressed sensing is selected to compress a n image in order to preserve more details of the image in the condition of the same compression ratio.Then,wavelet based contourle t transform (WBCT) is chosen as the sparse representation to extract more direction information for the demand of completely maintaining t he texture and boundaries of the reconstructed image. Finally,an ROI boundary smoothing method is provided to effectively improve the subjective visual quality of the reconstructed image, for overcoming the deficiency of over differentiating between ROI and backgrou nd of the reconstructed image and loosing excessive ROI-related information.Experimental results indicate that the total c ompression ratio falls to 0.18in the case that the ROI compression rate is as high as 0.4,which both protects the ROI and signifi cantly improves the overall compression ratio.The algorithm can enha nce the efficiency of wireless image transmission,supplying subsequent detection and recognition task with high quality images.