OAK

Compressive Sensing CMOS Image Sensor

Metadata Downloads
Author(s)
Hyunkeun Lee
Type
Thesis
Degree
Doctor
Department
대학원 전기전자컴퓨터공학부
Advisor
Lee, Byung-geun
Abstract
For high-resolution and high-frame-rate CMOS image sensors (CISs), the analog to digital (A/D) conversion of pixel values becomes a limiting factor because it requires a large number of high-resolution analog to digital converters (ADCs) operating at high speeds and consuming moderate power. Compressive sensing (CS) can allow signal recovery from fewer samples than required by the Shannon-Nyquist sampling theorem by exploiting the sparsity of a signal. This approach reduces the number of A/D conversions performed and, subsequently, the total conversion time and power consumption.
In this dissertation, two CS-CISs that perform CS image encoding are presented. The conversion rate and the frame rate of the sensor are increased by using a high-order sigma-delta (ΣΔ) ADC to obtain linear measurements of selected pixel values for CS encoding. Image distortion caused by the nonconstant weight function of high-order ΣΔ ADCs (e.g., 2nd- or 3rd-order) is eliminated by using the proposed sampling techniques. The sampling techniques make the effective weights of the inputs to the ADC equal. In addition, hardware-efficient CS encoding schemes help to reduce hardware complexity and control compression ratio without hardware modifications.
The prototype designs of a 160×160-pixel CS-CIS with 2nd-order ΣΔ ADCs and a 256×256-pixel CS-CIS with 3rd-order ΣΔ ADCs implemented in 0.11-μm 1P4M CIS process successfully demonstrate the proposed techniques and encoding schemes. The first CIS achieves a readout noise of 29.7 e-rms and a dynamic range of 62.08 dB with a power consumption of 37.1 mW. The sensor can operate at 240 fps to 1200 fps as the compression ratio (CR) decreases from 1 to 1/5. The second CIS achieves a readout noise of 2.63 e-rms and a dynamic range of 67.96 dB with a power consumption of 56.38 mW. The frame rate of the second CIS can vary from 781 fps to 6248 fps as the CR decreases from 1 to 1/8. The resulting figure of merit (FoM), defined as, FoM = (power x noise)/(# of pixels x frame rate) , is 0.375 e-∙nJ.
URI
https://scholar.gist.ac.kr/handle/local/33006
Fulltext
http://gist.dcollection.net/common/orgView/200000908897
공개 및 라이선스
  • 공개 구분공개
파일 목록
  • 관련 파일이 존재하지 않습니다.

Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.