One of challenges encountered in information bit hiding is the reliability of information bit detection. This paper addresses the issue and presents an algorithm in the discrete cosing transform (DCT) domain with a communication theory approach. It embeds information bits (first) in the DC and (then in the) low-frequency AC coefficients. To extract the hidden information bits from a possibly corrupted marked image with a low error probability, we model information hiding as a digital communication problem and apply Bose-Chaudhuri-Hocquenghen channel coding with soft-decision decoding based on matched filtering. The robustness of the hidden bits has been tested with StirMark. The experimental results demonstrate that the embedded information bits are perceptually transparent and can successfully resist common signal processing procedures, jitter attack, aspect ratio variation, scaling change, small angle rotation, small amount cropping, and JPEG compression with quality factor as low as 10. Compared with some information hiding algorithms reported in the literature, it appears that the hidden information bits with the proposed approach are relatively more robust. While the approach is presented for gray level images, it can also be applied to color images and video sequences.