Perceptual Quality Assessment of Wireless Video ... - ResearchGate

eference for the calculation of the distorted imagequality. Therefore, these approaches are not suitable forwireless communication purposes as the original imagewould typically not be available at the receiver. Instead,reduced-reference (RR) image quality metrics can beused which shall be based on algorithms that extractfeatures such as structural information from the originalimage at the transmitting end. The feature data maythen be sent over the channel along with the image. Atthe receiver, the image related data is extracted and thefeatures of the received image are calculated. Given thefeatures of the transmitted and received image, a qualityassessment can be performed.In view of the above arguments, the favorable perceptualvideo quality assessment shall be based on suchan RR image quality metric. This approach finds itssupport in the fact that MJ2 videos consist of frameswhich are entirely intra-frame coded. This means thatthere are no dependencies between consecutive frames.Therewith, there are no temporal artifacts introducedthrough neither the MJ2 source coding nor the wirelesschannel. As such, the quality of each video frame canbe evaluated independently from its predecessors andsuccessors using suitable image quality metrics.The availability of the quality measure of eachMJ2 video frame may be applied for link adaptationand resource management algorithms to adapt systemparameters such that a satisfactory perceived qualityis delivered to the end user. The block diagram ofsuch an application scenario is presented in Fig. 1.The features of each frame are calculated in the pixeldomain of the uncompressed video frame. The resultingdata is then concatenated with the data stream of thevideo frame. Together they are sent over the channel.At the receiver, the data representing the features isextracted. After MJ2 source decoding the features of thereceived video frames are calculated and used, togetherwith the features of the sent video frames, for thequality assessment. On the grounds of this assessmenta decision can be deduced for the adaptation of systemparameters.2.1 Hybrid Image Quality MetricAs a reduced-reference metric, HIQM [9] extracts thefeatures of the video frames on both the transmitterand receiver. The quality evaluation is composed ofthe outcomes from different image feature extractionalgorithms such as blocking [10], [11], blur [12], imageactivity [13], and intensity masking [14]. Due to thelimited bandwidth of the wireless channel it is anobjective to keep the resulting overhead needed torepresent the video frame features as low as possible.Therefore, the overall perceptual quality measure shallbe calculated as a weighted sum of the extractedfeatures to be represented by a single number. Thisnumber can be concatenated with the data stream ofeach transmitted video frame without creating too muchTABLE IARTIFACT EVALUATION.Feature/Artifact Metric Algorithm Weight ValueBlocking f 1 [11] w 1 0.77Blur f 2 [12] w 2 0.35Edge-based activity f 3 [13] w 3 0.61Gradient-based activity f 4 [13] w 4 0.16Intensity masking f 5 [14] w 5 0.35overhead. Specifically, the proposed metric is given byHIQM =5∑w i · f i (1)i=1where w i denotes the weight of the respective imagefeature f i , i = 1, 2, 3, 4, 5. It is noted that the followingrelationships have been used:f 1 Blocking metricf 2 Blur metricf 3 Edge-based image activity metricf 4 Gradient-based image activity metricf 5 Intensity masking metricIn order to obtain the values of the aforementionedweights, subject quality tests have been conducted atthe Department of Signal Processing of the BlekingeInstitute of Technology and an analysis of the resultshas been performed for the individual artifacts. The testwas performed using the Double Stimulus ContinuousQuality Scale (DSCQS) methodology, specified in ITU-R Recommendation BT.500-11 [15]. A total of 30people had to vote for the perceived quality of boththe transmitted and received set of 40 images. Theresponses of the test subjects are captured by the respectivePearson correlation coefficients. Accordingly,the magnitudes of these correlation coefficients areselected as the weights by which the individual artifactscontribute to the overall HIQM value (see Table I). Thefinal quality measure of an MJ2 encoded video frame atthe receiver may then be represented by the magnitudeof the difference between the feature measure of thetransmitted and the received frame∆ HIQM (i) = |HIQM T (i) − HIQM R (i)| (2)where i denotes the i th frame within the transmitted (T )and the received (R) video stream. The total length ofthe time-varying HIQM related quality value may berepresented by 17 bits (1 bit for the sign, 8 bits for theinteger in the range 0-255, 4 bits for each the 1 st andthe 2 nd decimal).Several other image quality metrics have been proposedin recent years. For comparison purposes we willconsider in the sequel two metrics for which the sourcecode has actually been made available to the public.