A metadata-free pure soft broadcast scheme for image and video transmission

Abstract

The SoftCast, as an analog-like transmission scheme, has been proposed to provide graceful video quality degradation for wireless communication scenarios where channel condition usually varies dynamically and unpredictably. It attempts to eliminate the cliff effect that conventional communication systems usually suffer from. The unequal power allocation minimizes the total distortion by optimally scaling the transform coefficients according to their statistics. However, the sender has to transmit considerable amount of metadata over a digital channel with strong channel codes to assist in decoding at the receiver. And the transmission will still break down completely if the metadata are corrupted. Hence, SoftCast can just alleviate the cliff effect. To eliminate the cliff effect while maintaining the image and video quality, a feasible metadata-free pure soft broadcast scheme for image and video transmission is proposed in this paper. Specifically, we apply an effective chunk-based blind estimation method to reconstruct the original signal from the received noisy signal without using power scaling factors. Thus, our scheme can eliminate the cliff effect. In addition, since smaller chunks can be used, better performance than SoftCast has been achieved. Experimental results show that our scheme outperforms SoftCast by about 4dB and provides graceful quality degradation for a wide channel signal to noise ratio range.