Frame alignment and insertion algorithm for video stream over mobile network

Abstract

Video stream over mobile network has become popular service influencing trend and lifestyle in society. The condition is highly correlated with customer satisfaction that needs to be managed by service stakeholders. One of the efforts is to evaluate their service quality that can be represented by measuring quality of video that they stream to the customer. Peak Signal to Noise (PSNR) is a dominant method to measure video quality that has been widely used. However, PSNR concept cannot be used for video that streamed over wireless and mobile network. It is due to occurrence of packet loss that is inherent in the network and results to frame loss. To overcome this shortcoming, this research proposes an algorithm called frame alignment algorithm to locate frame loss position and do measurement on pair of corresponding frames between frame in reference video and frame in streamed video. This research also enhances the first algorithm accuracy by inserting adjacent frame to the frame loss position, frame insertion algorithm. Simulation of video stream has been conducted to generate video test material, and the proposed algorithms are used to measure quality of the video test material. Performance of the the proposed algorithms are evaluated by benchmarking the experiment result towards the conventional PSNR and the equivalent method, Modified-PSNR (MPSNR) based on Pearson product-Moment Correlation Coefficient (PMCC) value. Based on the result, frame alignment algorithm achieves 0.85 in terms of PMCC value that overcomes inaccuracy of the conventional PSNR that only results 0.77. Frame alignment algorithm also has better performance than MPSNR method that only reaches 0.84. Besides, the investigation on frame insertion algorithm can also enhance accuracy with PMCC value of 0.86. From this result, the proposed algorithms are potential to be used as an alternative method in measuring streamed video quality that still keeps simplicity of PSNR concept but with better accuracy