| 摘要 |
A method of reconstructing pixel values of a video frame comprises: receiving a signal from a communication channel, delivering video packets comprising pixel values possibly corrupted with errors; associating confidence levels with packet pixel values; and reconstructing pixel values for display, weighted by their associated confidence levels. One of plural possible pixel reconstruction strategies may be selected, depending on the confidence level associated with at least one pixel value in the pixel set. Confidence levels â for example, based on signal-to-noise (SNR) ratios or inverse residual bit error rate (BER) of luminance and/or chrominance data â may be compared to successive, plural thresholds; based on comparison with multiple thresholds, pixels may be corrected and reconstructed by: simply keeping the pixel unchanged, S615 (SNR sufficiently high); maximum ratio combining (MRC) of all pixel values, S616; pixel averaging, S617, based on absolute values of pixel differences, S607; or by duplicating the pixel with the highest confidence (least corrupted, e.g., maximum SNR), S618. Alternatively, pixel values may be obtained by MRC combination of all pixels in a set if a cyclic redundancy check (CRC) indicates errors in a video packet and a confidence level above a certain threshold. |
