Abstract
To realize low delay live video streaming system over lossy network, each captured frame must be divided into block images, and these images must be encoded, packetized, and sent selectively. To improve the quality of reproduced frames at the receiver side, the selecting method of the block images must be sophisticated. This paper proposed an optimal selecting method of block images. The method consists of three phases based on the formulated two optimization problems. Firstly, a frequency analysis method is applied to the recent luminance value for each block area. Secondly, sendable block ratio is optimally distributed among block areas based on the result of frequency analysis to satisfy the available bandwidth condition. Finally, the optimal timings of sending are planned for each block area to minimize the error of reproduced image under the condition of assigned sendable block ratio. Some experiments were conducted to evaluate the proposed method using three frequently used reference video sequences. The results of comparison with the conventional method shows that the proposed method improved the mean squared error (MSE) of reproduced image up to 64%. Another results shows that the robustness of the proposed method against packet loss can be changed by modifying the parameter of the optimization problem.
