Multiple Description Coding Based on Convolutional Auto-Encoder

Abstract

Deep learning, such as convolutional neural networks, has been achieved great success in image processing, computer vision task, and image compression, and has achieved better performance. This paper designs a multiple description coding frameworks based on symmetric convolutional auto-encoder, which can achieve high-quality image reconstruction. First, the image is input into the convolutional auto-encoder, and the extracted features are obtained. Then, the extracted features are encoded by the multiple description coding and split into two descriptions for transmission to the decoder. We can get the side information by the side decoder and the central information by the central decoder. Finally, the side information and the central information are deconvolved by convolutional auto-encoder. The experimental results validate that the proposed scheme outperforms the state-of-the-art methods. INDEX TERMS Convolutional auto-encoder (CAE), multiple description coding (MDC), predictive coding, quality metric.

Deep Multiple Description Coding by Learning Scalar Quantization

“…When the artificial neural network represents an input image as the feature tensors, a pair of scalar quantizers is learned to quantize the feature tensors for diversified multiple description generation. Recently, a convolutional autoencoder-based multiple description coding method [32] has extracts features by learning, which improves image coding efficiency. However, this method suffers from severe coding artifacts similarly to conventional MDC approaches.…”

Multiple Description Coding for Enhancing Outage and Video Performance Over Relay-Assisted Cognitive Radio Networks

“…In [4], MDC was used with the K-singular value decomposition (K-SVD) algorithm to achieve the sparse transform for reconstruction accuracy in image coding. The study in [5] proposed an MDC framework with auto-encoder to produce high-quality image reconstruction. As in video coding, [6] introduced MDC with a scalable coding technique (MDSC) for adaptive video streaming applications over cellular networks.…”

Deep Spatio-Temporal Representation and Ensemble Classification for Attention Deficit/Hyperactivity Disorder

“…CNN has advantages in image processing due to the ability to extract the spatial information hidden in the image. It is instinctively assumed that CNNs can work better than other autoencoders when constructing an encoder and decoder network, hence why the convolutional autoencoder (CAE) is generated [21].…”