- 學位論文
在一個無線平台上傳送經壓縮之膠囊內視鏡影像的效能評估
Performance Evaluation of Transmitting Compressed Capsule-Endoscope Images over A Wireless Platform
摘要
傳統內視鏡檢查必須吞入長長的胃鏡，除造成患者不適，產生恐懼感外，還有檢查死角，因此近年來發展出膠囊內視鏡，讓患者吞入一個如膠囊般大小的微型攝影機，並將6-8小時之久的腸道內影像傳送至病患身上的接收機上，再將影像自接收機以有線的方式傳送到主電腦端供醫生判斷病徵。若進入接收端的影像能立即以無線傳輸的方式傳到主電腦上，則醫生可爭取時效提早診斷。此外，由於資料量甚大及影像保真的考量，所以近乎無失真壓縮有其必要性。 因此，本文針對膠囊內視鏡，提出一套近乎無失真壓縮的無線傳輸系統。在病患端接收機上，首先將收到的膠囊內視鏡影像進行近乎無失真壓縮，再由以IEEE 802.11g與智慧型天線構成的無線傳輸平台將壓縮後的影像資料傳送至主電腦端，在主電腦端解碼還原影像後供醫生在第一時間進行檢視，待病患繳回接收機後可進一步將殘留其中的保真資訊加入至上述還原影像，以獲得近乎完美品質的重建影像。 實驗結果顯示影像經過近乎無失真壓縮還原後仍有一定的影像重建品質，且壓縮率可達5以上。主電腦端所還原的近乎無失真影像在SNR = -2 dB時，PSNR值平均約為40 dB，若加入接收機中的保真資訊則平均PSNR值可提升至80-90 dB左右，接近完全無失真。因此我們預期這樣的系統將可以有不錯的效能，達到醫生提早觀看影像的目的。
並列摘要
In the traditional endoscopy examination, a patient must swallow a long tube, which makes the patient feel uncomfortable and causes fear. Besides, it has inspection blind spots. Therefore, it motivates the recent development of capsule endoscopy, where a patient swallows a capsule-like micro-camera and transmits an intestinal image sequence in 6-8 hours to a receiver carried by the patient. Then these images are transmitted to a desktop computer in a wired connection so that a doctor can use them for diagnosis purpose. If we transmit the received images in the receiver with wireless transmission to the desktop computer, then the doctor can start the diagnosis much earlier. Furthermore, due to the consideration of image fidelity and massive quantity of image data, near-lossless compression is necessary. Thus, this thesis proposes a near-lossless compression and wireless transmission system for capsule endoscopes. First, at the receiver carried by the patient, we use a near-lossless compression algorithm to compress capsule endoscope images, then these compressed images data are transmitted to the desktop computer via a wireless transmission platform comprising of IEEE 802.11g and smart antenna systems. At the desktop computer, compressed images are decoded and reconstructed, the resulting images can be reviewed by doctors at early time. When the patient returns the receiver back, the fidelity-preserving data stored in the receiver can be added to the previously reconstructed images, resulting in almost perfect quality of image reconstruction. Experimental results show that after near-lossless compression, the reconstructed image still has fine quality and the compression ratio can reach 5. At the desktop computer, the reconstructed and near-lossless image has about 40 dB PSNR value on the average when SNR = -2 dB. After adding the fidelity-preserving data, the average PSNR value can reach about 80-90 dB, which is almost lossless. Thus, our system is expected to have good enough performance that can meet the goal of early image viewing for doctors.