The Research Of The Multiple Description Coding Of Image

Because of the development of multimedia technology, the transmission rate of the network channel becomes faster and faster, and in the high-speed network environment, the traditional packet-based data transfer is usually faced with the problems such as packet loss and bit error, so we pay more and more attention to the accuracy of the data transfer. Multiple Description Coding is an arithmetic that has strong rubstness. It divides the source signal into several streams. Each stream is a description and any independent descriptions are able to recover the old signal in receptible quality. The more descriptions received the better quality recovered. Multiple description coding is widely used in video, images, and multimedia signal processing and gets very good results. This paper researchs on multiple description image coding mainly in the following two aspects:1. The improvement of the MDSQ algorithm by the frame-expansionMultiple Description Coding based on scalar quantization is first proposed by Vaishampayan. It quantifies the original signal by a complex lable function and then gets two less-detail streams. When we face the poor channel quality, the loss of any stream will bring a significant effectt to the resule. So we use the normalized tight frame which has the strong robustness to expand the original signal from the two- dimension space to the multi-dimension space. Then we can get multiple streams which can recover the signal at the received end. In the simulation experiment, we use the 4×2 harmonic tight frame to expand the image to four description instead of two descriptions. The result shows that the frame-expanded MDSQ algorithm has a higher PNSR than the original algorithm in the small error probability channel.2. Multiple Description Coding based on the nonlinear transformation of the region of interestThe lose of the multiple description will definitely affect the recovery of the image ,but not all the parts of the image have the same importance to us. So we need to pay more attention to the region of interest of the image. So we presents a non-linear transformation multiple description coding. It use a special non-linear function to transform the image, so that the regions of interest of the image can be expanded and then samples the expanded-image to form the multiple description coding. The simulation results show that when the image-descriptions are lost in the error-prone channels, this algorithm will get a higher PSNR of the ROI than the ordinary algorithm.The new algorithm is able to get satisfactory results in the different bit rate.