Research On Quality Assessment And Related Issues In Compressive Sensing Video Transmission

In the current era of substantial growth on information technology,image and video as the carrier of the service application have already shown the significance in information communications and obtained widespread attention.The predictive video coding method like H.264 as the representative of mature and efficient algorithms is the basis of image and video processing.However,due to the requirement of a huge amount of calculations and resulting in diffusion of channel errors,the predictive video coding is inappropriate to some wireless video application scenarios with limited resources and low requirements,such as participation sensing,wireless multimedia sensor network,etc.Compressive sensing theory provides a new way to solve the above problems.It can not only break through the limitation of Nyquist frequency in sampling rate,but also achieve a certain degree of compression by dimension reduction in observation.As implementation of sampling and compression by merging two process,it avoids the wasted resource consumption caused by traditional sampling before compression way.Compressive sensing is a kind of global information acquisition mode in place of the traditional local signal acquisition,its measurements,which are acquired from linear random projection and bearing the same importance of information,benefits to construct a simple and effective anti channel error scheme.Combining the CS theory and the existing video technology,which can effectively reduce the stress of high speed sampling and the consumption of data acquisition,encoding,processing and transmission,will further promote the information field forward.On the basis of compressive sensing theory,and taking image & video signal as main research object,the research of dissertation focuses on three key issues about CS video transmission quality: CS video quality assessment,CS video efficient rate and CS video error control.The main contents and innovations of this dissertation are summarized as follows:First of all,in aspect of CS video quality assessment,a reduced reference CS video quality assessment based on redundant measurements method and a hierarchical objective CS video quality assessment method which meets the requirements of subjective quality are proposed.The former,in the form of objective,implements the CS video recovery quality assessment at low cost with a few additional redundant measurements as reduced reference.The quality index has a strong correlation with PSNR values.Meanwhile,according to the acquired video quality index,the measurement rate adaptive adjustment scheme with quality index feedback applied to CS video can be realized.Its overall quality is effectively improved compared to fixed rate of CS video transmission.The latter's model describes influence on CS video quality with transmission network parameters from the measurement level,stream level and packet level,respectively.And it can provide different level CS video quality index according to application needs.Meanwhile,all coefficients in the models are obtained by subjective CS video quality data analysis regression,so the quality index from model has human perception characteristics and reflects the requirements of subjective quality.The quality index from model has strong correlation with subjective CS video quality data from verification experiments.Secondly,in aspect of rate efficiency improvement,a measurements truncated threshold quantization method for CS image is proposed.In basis of characteristics of measurements: global projection and properties of approximate Gaussian distribution,by means of measurements value range truncation,it achieves the lower quantization error under a certain number of quantization levels due to lower measurements value range,and then promotes the rate distortion performance of measurements quantization consequently.The recovery image quality of this method in suitable truncation parameters outperforms that of direct uniform scalar quantization method and the block CS image DPCM method.Finally,in aspect of CS image/video error control,a saliency-based scalable coding error control method and a detection & deletion error control method based on monobit parity check for CS video measurement are proposed.The former,aiming at problem of the low reliability of single path transmission of large data like image signal,implements the scalable coding,which is asymmetric channel error resilience,by multipath diversity and image saliency analysis technology.The rate distortion performance of this method is better than that of the traditional no-saliency method,and better than that of the CS multiple description coding method in the packet loss environment.The latter,considering the characteristics of measurements in CS video transmission,implements the error control in CS video transmission by error detection & deletion based on monobit parity check.This method is simple and the parameters can be adjusted according to the channel state.Compared with commonly used error control methods,it has obvious advantages in bit rate and computational complexity,and achieves performance of BCH codes and RCPC codes in different bit error rate channel conditions.