Video Quality Evaluation using NR metric with Detection of Frame Dips and Drops

Abstract: Wireless technology has become the most interesting field in the area of communications and networking. Video transmission plays a major role in this area. Data transmission via wireless communications is a major task, especially to send the data without loss of packets. The digital video decoder produces dropped or repeated frames from the encoded video due to packet loss. Brief frame drops of one frame duration are referred as dips. This research work investigates how to identify these frame dips and drops and also to measure the quality of the video which contain frame dips and frame drops. A dynamic threshold estimator is implemented to identify these frame dips. The qualitative analysis of a video is performed using a No Reference metric. In this metric the quality assessment can be achieved by identification of frame dips and drops and calculating Modified Fraction of Dropped Frames. Dips and abrupt temporal variation occurring at the end of frame freezing event of a video fluidity impairment caused by frame dropping will be taken into account using this metric. The parameters which are generated from Structural Similarity Index Metric (SSIM) analysis are efficient enough to compare the video quality and estimate the metric performance objectively. Our proposed metric has predicted the SSIM with a Pearson correlation coefficient of 0.99 in test1, 0.92 in test2 and 0.99 in test3. Therefore in terms of prediction accuracy the proposed metric has a good correlation with the deployed metric and obtained robustness to our approach.