DESIGN AND IMPLEMENTATION OF AN EFFICIENT IMAGE COMPRESSOR FOR WIRELESS CAPSULE ENDOSCOPY
Mostafa, Atahar Kamal
Capsule endoscope (CE) is a diagnosis tool for gastrointestinal (GI) diseases. Area and power are the two important parameters for the components used in CE. To optimize these two parameters, an efficient image compressor is desired. The mage compressor should be able to sufficiently compress the captured images to save transmission power, retain reconstruction quality for accurate diagnosis and consumes small physical area. To meet all of the above mentioned conditions, we have studied several transform coding based lossy compression algorithms in this thesis. The core computation tool of these compressors is the Discrete Cosine Transform (DCT) kernel. The DCT accumulates the distributed energy of an image in a small centralized area and supports more compression with non-significant quality degradation. The conventional DCT requires complex floating point multiplication, which is not feasible for wireless capsule endoscopy (WCE) application because of its high implementation cost. So, an integer version of the DCT, known as iDCT, is used in this work. Several low complexity iDCTs along with different color space converters (such as, YUV, YEF, YCgCo) were combined to obtain the desired compression level. At the end a quantization stage is used in the proposed algorithm to achieve further compression. We have analyzed the endoscopic images and based on their properties, three quantization matrix sets have been proposed for three color planes. The algorithms are verified at both software (using MATLAB) and hardware (using HDL Verilog coding) levels. In the end, the performance of all the proposed schemes has been evaluated for optimal operation in WCE application.
DegreeMaster of Science (M.Sc.)
DepartmentElectrical and Computer Engineering
SupervisorWahid, Khan A.; Ko, Seok-Bum
CommitteeGokaraju, Ramakrishna; Dinh, Anh V.; Yang, Qiaoqin
Copyright DateJune 2012
Wireless capsule endoscopy, Image compressor, Transform coding, Endoscopic image.