A 2PJ/Pixel/Direction MIMO Processing Based CMOS Image Sensor for Omnidirectional Local Binary Pattern Extraction and Edge Detection

This paper presents an energy-efficient multi-mode CMOS image sensor featuring omnidirectional local binary pattern extraction (LBP-E), edge detection (ED) and normal imaging. A mixed-signal 4-pixel simultaneous group computation (GC) scheme is developed to extract complete 8-direction LBP and edge....

Full description

Saved in:
Bibliographic Details
Published in2018 IEEE Symposium on VLSI Circuits pp. 247 - 248
Main Authors Xiaopeng Zhong, Qian Yu, Bermak, Amine, Chi-Ying Tsui, Law, May-Kay
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.06.2018
Subjects
CMOS image sensors
Feature extraction
Histograms
Image edge detection
MIMO communication
Prototypes
Online AccessGet full text
DOI10.1109/VLSIC.2018.8502214

Cover

More Information
Summary:This paper presents an energy-efficient multi-mode CMOS image sensor featuring omnidirectional local binary pattern extraction (LBP-E), edge detection (ED) and normal imaging. A mixed-signal 4-pixel simultaneous group computation (GC) scheme is developed to extract complete 8-direction LBP and edge. High-speed and energy-efficient column-parallel GC is achieved with the proposed group-switchable multi-input multi-output (MIMO) comparator. The reconfigurable mixed-signal processing circuits ensure multi-mode operations with minimum area overhead. Fabricated in 0.18μm CMOS, the prototype sensor consumes 6.5μW and 12.7μW at 30fps for full 8-direction LBP-E and ED, achieving the state-of-the-art FoMs of 2.0 and 3.9 pJ/pixel/direction, respectively. Measurement results also demonstrate high-accuracy on-chip LBP-E, with a histogram similarity of up to 97.5% compared to the DSP one.
DOI:10.1109/VLSIC.2018.8502214