Biography
Enrollment Date: 2009
Graduation Date：2012
Degree：M.S.
Defense Date：2012.05.25
Advisors：Xiang Xie
Department：Institute of Microelectronics，Tsinghua University
Title of Dissertation/Thesis：Research on the Image Compression Technology of Wireless Endoscopic Micro-Ball System
Abstract:
The wireless endoscopy system is a device to wirelessly detect the human digestive tract diseases. The capsule endoscopy contains only one or two cameras which will take an omission problem of diseases. A micro-ball endoscopy system proposing by our team which contains six cameras will solve the omission problem. The micro-ball endoscopy system need capture larger number of images than capsule endoscopy. To extend the work time, the micro-ball system doesn’t send the image out in real time but store them in memory. This method will need larger transmission bandwidth and smaller transmission power. The compression way provided by capsule endoscopy already cannot meet the demands of the micro-ball system. Based on this, this thesis designs and implements a low power high compression rate and high image quality image compression method for the micro-ball endoscopy system.
The compressor is designed using JPEG 4x4 algorithm based on Bayer format image. The digital circuits are designed by verilog language and simulated by modelsim software. The design uses several low-power and low-area techniques. Though verification and layout, the compressor is implemented by UMC18. The test results prove that this compressor can compress 240x240 and 480x480 resolution images and can compress 1~6 frames per second. The area of this compressor is about 0.46 mm2 and the power is 0.9mW@2fps1.8V.
To the blocking artifact bringing by high compression ratio based on Bayer format JPEG algorithm, this thesis provides a novel random spatial adjust method based on pseudo random number generator. This method can greatly remove the blocking artifact and improve the subjective effect of the compressed images. Though random spatial adjust, the compression ratio can improve from 7~8:1 to 10:1 and the compressed images can still achieve approximate subjective effect with the original images.
In this thesis, the digital circuits of micro-ball system are verified by verilog language and C language. The verification platform is designed using reusable modules and the results of the design are compared automatically with the correct results which significantly improve the effective of the verification. The digital circuits of the micro-ball system is verified by FPGA and implemented by UMC18. Test results prove that the micro-ball system can work properly which can further prove the verification platform is correct and usable.