A Hybrid Logarithmic–Inverse Counting CMOS Image Sensor for High Dynamic Range Imaging
- 발행기관 서강대학교 일반대학원
- 지도교수 Seong-Jin Kim
- 발행년도 2026
- 학위수여년월 2026. 2
- 학위명 석사
- 학과 및 전공 일반대학원 반도체공학과
- 실제URI http://www.dcollection.net/handler/sogang/000000082886
- UCI I804:11029-000000082886
- 본문언어 영어
- 저작권 논문은 저작권에 의해 보호받습니다.
초록(요약문)
Recent image sensors must cope with extreme illumination variations in applications such as autonomous driving, security, mobile imaging, and industrial machine vision. As a result, sensor technologies capable of providing a high dynamic range (HDR) have become essential. When HDR performance is insufficient, pixel saturation in bright regions leads to highlight clipping, while the signal-to-noise ratio (SNR) in dark regions degrades sharply, causing loss of edge and texture information. Such issues not only deteriorate image quality but also negatively affect recognition performance in sensors such as vision sensors. Therefore, securing sufficient dynamic range at the sensor level is a key factor in determining overall system stability and accuracy. To address this need, many HDR sensors have adopted the Time-to-Saturation (TTS) method. TTS measures the time until a pixel saturates, where strong illumination results in short saturation times and weak illumination results in long saturation times. This allows a wide dynamic range to be obtained using relatively simple circuitry. However, the TTS method has two fundamental limitations. First, achieving a higher dynamic range requires continuously increasing the number of bits used to store the Time Code(TC). Because the dynamic range is directly tied to time-code resolution, widening the DR necessitates a larger time-code bit-width. This increase leads to larger in-pixel memory, expanded circuit area, and higher power consumption, which significantly limits practicality in high-resolution or compact mobile devices. Second, the amount of information represented by each time code decreases sharply as illumination becomes stronger. In high-illumination conditions, the saturation time becomes extremely short, and although the time-code intervals may be densely arranged, the amount of representable brightness variation per code becomes very small. As a result, the reconstructed image exhibits banding effects— where gradation collapses or striping artifacts appear in bright regions. Consequently, conventional TTS alone cannot provide both a wide dynamic range and natural image quality. To overcome these issues, this work proposes a new HDR encoding architecture that combines inverse time-code counting with logarithmic time-code counting. Inverse counting refines the coarse time codes obtained from the TTS process by performing finer measurements in regions where the image becomes smoothed or degraded due to strong illumination. In other words, it re-examines under-resolved time-code segments to mitigate banding effects and recover detail in high-light regions. Meanwhile, logarithmic time-code counting enables representation of a much wider illumination range with minimal increases in bit-width. By applying logarithmic time-domain encoding, significantly broader dynamic range can be expressed compared to conventional linear TTS, even with only a small memory and circuit overhead. This makes it possible to target ultra-high dynamic ranges exceeding 120 dB. As a result, the proposed inverse + logarithmic time-counting architecture achieved more than 90 dB improvement in dynamic range compared to the conventional TTS method, while requiring significantly fewer bits. Furthermore, by analytically supplementing coarse time codes, the proposed method achieved over a 512× enhancement in effective temporal resolution, mitigating banding artifacts in bright regions while preserving fine detail in low-light regions. This allows stable and natural HDR imaging across the entire illumination range Keyword: High Dynamic Range (HDR), Time-to-Saturation (TTS), signal-to-noise ratio (SNR), Time Code(TC), Inverse time-code counting, Logarithmic time-code counting
more목차
CHAPTER 1 Introduction 8`
Chapter 1.1: Applications requiring High dynamic range imaging 8
Chapter 1.2: CMOS Image sensor 10
Chapter 1.3: Definition of Dynamic range 13
Chapter 1.4: Organization of thesis 15
CHAPTER 2 Preious HDR Scheme 17
Chapter 2.1: Multi-Exposure 17
Chapter 2.2: Dual Conversion Gain 19
Chapter 2.3: Time-To-Saturation 21
CHAPTER 3 Proposed HDR Imaging Architecture Based on an Enhanced TTS Method 23
Chapter 3.1: Problem of time to saturation – Banding effect 23
Chapter 3.2: Problem of time to saturation – In pixel memory 26
Chapter 3.3: Inveres Time Code Counting 27
Chapter 3.4: Logarithmic Time Code Counting 28
Chapter 3.5: Summary of proposed idea 30
CHAPTER 4 Circuit Implementation of the Proposed Sensor Architecture 31
Chapter 4.1: Pixel Architecture 31
Chapter 4.2: Peri-ciruict Architecture 36
Chapter 4.3: ADC & CDS circuit Architecture 38
CHAPTER 5 Measurement Result 40
Chapter 5.1: DR measurement 40
Chapter 5.2: Imaging result 41
CHAPTER 6 Conclusion 42
References 44
Acknowledgements 46
