Image transmission over error-prone channels: sigma filtering and multiple description objectives

This work considers a fundamental problem of image transmission over error-prone channels. The impairments that we target are long burst errors and transient channel failures, as would occur in the wireless network because of an obstruction in the transmission path, or severe traffic congestion in a packet network such as the Internet. These impairments cause many packets to be dropped. In contrast to the traditional schemes that first introduce some amount of redundancy among descriptions, and then use this correlation for combating channel impairments, we investigate a new scheme to tackle this problem. The proposed scheme is to capture the most important visual features of a given image, extracted by sigma filtering preprocessing, into independent equal length packets. At the receiver side, we adopt a smart interpolation-based technique. The reconstruction quality at the receiver depends only on the number of packets received, but is independent of the place from where they were cropped. The preliminary results on standard test images show that our reconstructed image is very pleasant to human eyes, and achieves reasonable PSNR values, while at the same time using a lower bit rate than other reported traditional schemes based on multiple description coding