Haechul Choi

1. Introduction Screen content refers to video content generated without camera capture. Typical applications include remote desktops, Wi-Fi displays, and cloud gaming, which deliver computer-generated text, graphics, and animations [1]. However, screen content has unique challenges in compression using video coding solutions, as it is free from the sensor noise that is typical of camera-captured videos and often includes large homogeneous areas, repeated patterns, and identical blocks [2].

4.3.3. Visual Comparison A tracker must be able to handle various distractors to realize robust tracking in real-world applications. Figure 6 shows a visual comparison of the results obtained by the proposed method and other methods, i.e., MedianFlow, SiamRPN++, YOLOv4, and LC, on the test dataset. Note that all methods were initialized in the first frame for this evaluation. In the first row, Seq4 demonstrates that MedianFlow and SiamRPN++ can fail to track the target in out-of-view cases.