Selective error detection and error concealment for error-resilient wavelet-based image coding

This dissertation focuses on the transmission of images over channels with bit errors. As multimedia applications may not require perfect reconstruction of visual data, a satisfactory perceived quality may be sufficient to the end user. The present work proposes to perform error detection in the perceived domain according to a desired quality instead of detecting all bit errors, which allows a significant decrease in overhead for error control.; This dissertation introduces the concept of a similarity check function for error-resilient multimedia data transmission. The proposed similarity check function provides information about the effects of corrupted data on the quality of the reconstructed image. The degree of data corruption is measured by the similarity check function at the receiver, without explicit knowledge of the original source data.; This work also presents a selective error detection and error concealment (SEDEC) scheme based on the proposed similarity check function for wavelet-based image coders. It shows how non-perceptual and perceptual similarity check functions can be designed to significantly decrease the retransmission rate of corrupted data while maintaining very good visual quality of images transmitted over noisy channels.; The design of a similarity check function is presented for wavelet-based coders that results in independent corruption of coefficients when submitted to random bit errors, such as Trellis-Coded-Quantization-based coders, in addition to wavelet-based coders that incorporate entropy coding where a single bit error can corrupt subsequent coefficients in a block of data, such as the JPEG2000 standard.; Simulation results show that the designed similarity check functions can detect perceived visual errors, and can provide a good reconstructed visual quality with a minimal amount of retransmission.