A novel error correction method without overhead for corrupted JPEG images

This paper presents a novel error detection and correction methodology for corrupted coefficients caused by transmission of coded images based on orthogonal transforms over a noisy channel. This method is based on the orthogonal property of image transforms, such as the discrete cosine transform. A few reference pixel intensities of each image block are replaced by a predetermined intensity level prior to transmission. This allows the receiver to identify and correct the error pattern generated by the corruption of DCT coefficients. It is possible to correct t corrupted DCT coefficients in an image block by altering the value of 2t+1 pixel intensity levels to a predetermined value in each image block. After recovering the corrupted DCT coefficients, if necessary, and reconstructing the image, the original intensity level of each reference pixel is estimated by averaging the intensity levels of the adjacent pixels. The resulting image is indistinguishable from the original image when examined by a human. This algorithm does not require any channel overhead. An illustrative example is presented to demonstrate the performance of our algorithm.