Fellows 2009-2010
Pavel Khazron
Ph.D. (5th. year), Electrical Engineering
Year in graduate program: 5th. year (July 2005)
Research Topic: Image processing
Partner School: IS 383 Philippa Schuyler
Partner Teacher: Lindrick Outerbridge
Email Pavel
Research
Under the guidance of Prof. Ivan Selesnick, Pavel has developed a wavelet-based method of removing noise in digital video sequences that effectively combines existing results. The question of how to filter video along the time axis has been addressed in a way that does not result in temporal artifacts that arise from simple filtering. Motion detection and filtering is automatic based on estimates of motion and filtering parameters. In case of video with slowly varying content such as surveillance footage, more temporal filtering is performed, while in fast-moving video such as that observed in sports, the method relies more on spatial information to perform denoising. The developed denoising method can be applied in situations where efficiency is required: surveillance, video teleconferencing, and mobile applications.
Figure: Two denoised frames from video sequences with different characteristics. The frame on the top is from a slowly-varying video sequence, while the frame on the bottom is taken from a video with fast motion. In both cases, noise has been filtered out well while preserving spatial and temporal details such as cloth folds and player decals.
Integration in The Class Room
Pavel uses his research in middle school robotics and computer classes to promote several disciplines within mechatronics. He teaches computer programming concepts using the NXT Mindstorms graphical programming language. He emphasizes the logic behind computer programs through flow diagrams-a tool that has allowed 6th and 7th grade robotics students to better grasp programming fundamentals and write more flexible programs for the FLL competition. Motivated by the way cameras record video, Pavel is developing activities using the NXT light sensor that will teach students about the way images are acquired and stored. Pavel has also enhanced the popular mouse trap racing activity that uses the NXT acceleration sensor mounted atop a mouse trap racer, and allows students to compete not only on the basis of distance traveled but also other parameters: top velocity and acceleration, and smallest achieved deceleration. Integrating this sensor has made it possible for students to gain deeper understanding of the kinematic equations and the tradeoffs involved in real-world design of vehicles. In addition, Pavel is assisting the partner teacher in making use of magnetic and color sensors. For example, an interesting student project involving the use of the magnetic sensor has been a mine sweeper robot, while the color sensor has been demonstrated in class on a Rubik cube solver robot, which uses the color sensor to "see" and match the cube's six colors.
Figure: Mr. Outerbridge illustrates a microcontroller project to IS 383 students (top) and students work on the LEGO Mindstorm controller (bottom).
1 - AMPS Fellow; 2 - CBRI Fellow