Image Reconstruction Using Lapped Biorthogonal Transformation and Multiple Description Coding

Abstract

Multiple Description Coding (MDC) is one of the source coding methods in order to provide several independent and quantized streams of input data. Also, MDC is a useful source coding method for concealing error in lossy networks. In this paper, coded and quantized image streams are transmitted to the communication channel and any loss happens in the channel can be restored by averaging method. By using averaging method the average of received descriptions are taken and substituted with the lost descriptions. Input image is downsampled to 4 subimages. These 4 descriptions are mapped to transform domain by Lapped Biorthogonal Transform (LBT) besides are quantized and codded. Later, these descriptions are transmitted over the channel and once any link is lost at the receiver, by using averaging method, missing data is reconstructed. In the proposed method retransmission of the descriptions is not needed, which causes to have simplified method. The PSNR values are better and subjective evaluation element is clearer than the previous work, which causes to have higher throughput and less blur images. Furthermore, facial details of images are clearer and edges of the image details are sharper than the compared research.