Implementation of super-resolution imaging for small object tracking

Visual quality and the level of precision of object tracking are highly desirable in modern tracking. However, in reality, the object being tracked is not always clearly visible, so the tracking results obtained are less precise. One of the reasons is the small object. The super-resolution technique is one technique that can be done before the tracking process to obtain precise tracking results. Super-resolution imaging is a technique to transform a low-resolution image into a high-resolution image. Super-resolution single frame excellence is its fast computing time. In this research, Directional Bicubic Interpolation was proposed which has advantages in fast computing time and maintains edge sharpness. The process of object tracking uses a combination (called Hybrid) of the CamShift and Kalman Filter methods. The results of the trials resulted in an average tracking precision of 71% without super-resolution processes and 92% with super-resolution with the Camshift method. Whereas tracking using a combination method produces an average tracking precision of 73% without super-resolution process and 93% with super-resolution. So it can be concluded that the addition of the super-resolution process and the Kalman Filter prediction causes the increased precision of the object tracking results. The integration of the Kalman Filter into the Camshift method yields faster average tracking time compared with only using the latter method in the entire video.