Information for Paper ID 1172
Paper Information:
Paper Title: A 128×128 CMOS Image Sensor with In-Pixel Temporal-Occupancy Extraction for Single-Frame Blur-Based Fast Motion Estimation 
Student Contest: No 
Affiliation Type: Academia 
Keywords: CMOS Image Sensor, Temporal Occupancy Extraction, Motion Blur, Single-frame Motion Estimation, Fast Motion Estimation 
Abstract: This paper presents a 128x128 CMOS image sensor with in-pixel temporal-occupancy α-extraction for single-frame blur-based motion estimation. The proposed pixel measures intra-exposure occupancy over n sub-exposure cycles and encodes it as an analog voltage via comparator-controlled conditional charge injection, eliminating in-pixel digital counters. A column-level event detection circuit resolves direction ambiguity from the α-distribution using a pre-event signal for polarity determination. A single-ended, auto-zeroing column readout with self-calibration reuses the analog ramp path to suppress mismatch and process variation. The sensor supports multiplexed operation for 10-bit imaging and motion sensing. Fabricated in 180nm CMOS, the chip achieves 16 pixels/frame detectable motion at 100 fps, consuming 203.16 μW. Measured INL is < 0.45 LSB with non-uniformity < 0.96 LSB. The prototype demonstrates an iFoM of 124 for low-power edge vision applications. 
Track ID: 1 
Track Name: Analog Circuits & Systems 
Final Decision: Accept as Lecture 
Session Name: Analog Techniques (Lecture)