In the transmission of JPEG2000-coded remote sensing image (RSI), errors occurring in the codestream may result in the loss of the high-frequency information of RSI. A new method based on constraint net (CN) is proposed for error recovery of the high-frequency information. The CN of a wavelet coefficient block represents the relationships between each coefficient and its neighborhoods. The wavelet coefficient block has proved to be determined by its CN and surrounding neighboring coefficients. By using the correlation of inter-bands and the preservation property of edge feature and texture structure in high-resolution and low-resolution subbands, the CN of the damaged code-block that causes the loss of high-frequency information is computed. Together with the surrounding neighboring coefficients, the CN is used to recover the damaged code-block. Experimental results show the high-frequency information is well recovered and the quality of the reconstructed image is significantly improved.