A Datacompressive 15275Bit Loggradient Qvga Image Sensor With Multiscale Readout For Alwayson Object Detection

• doi: 10.1109/JSSC.2019.2937437
• title: A Data-Compressive 1.5/2.75-bit Log-Gradient QVGA Image Sensor With Multi-Scale Readout for Always-On Object Detection
• publisher: IEEE
• isbn:
• issn: 1558-173X
• rank: 686
• access_type: LOCKED
• content_type: Journals
• abstract: This article presents an application-optimized QVGA image sensor for low-power, always-on object detection using histograms of oriented gradients (HOG). In contrast to conventional CMOS imagers that feature linear and high-resolution ADCs, our readout scheme extracts logarithmic intensity gradients at 1.5 or 2.75 bits of resolution. This eliminates unnecessary illumination-related data and allows the HOG feature descriptors to be compressed by up to 25× relative to a conventional 8-bit readout. As a result, the digital backend-detector, which typically limits system efficiency, incurs less data movement and computation, leading to an estimated 3.3× energy reduction. The imager employs a column-parallel readout with analog cyclic-row buffers that also perform arbitrary-sized pixel-binning for multi-scale object detection. The log-digitization of pixel gradients is computed using a ratio-to-digital converter (RDC), which performs successive capacitive divisions to its input voltages. The prototype IC was fabricated in a 0.13-μm CIS process with standard 4-T 5-μm pixels and consumes 99 pJ/pixel. Experiments using a deformable parts model (DPM) detector for three object classes (persons, bicycles, and cars) indicate detection accuracies that are on par with conventional systems.
• article_number: 8844721
• pdf_url: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8844721
• html_url: https://ieeexplore.ieee.org/document/8844721/
• abstract_url: https://ieeexplore.ieee.org/document/8844721/
• publication_title: IEEE Journal of Solid-State Circuits
• conference_location:
• conference_dates:
• publication_number: 4
• is_number: 8880462
• publication_year: 2019
• publication_date: Nov. 2019
• start_page: 2932
• end_page: 2946
• citing_paper_count: 26
• citing_patent_count: 0
• download_count: 2736
• insert_date: 20190919
• index_terms:
  • ieee_terms:
    • Feature extraction
    • Lighting
    • Image coding
    • Object detection
    • Detectors
    • Histograms
    • Image sensors
  • author_terms:
    • CMOS image sensor
    • deformable parts model (DPM)
    • feature extraction
    • histogram of oriented gradient (HOG)
    • logarithmic gradients
    • object detection
    • pixel-binning
    • wake-up
  • dynamic_index_terms:
    • Object Detection
    • Image Sensor
    • Camera Sensor
    • Detection Accuracy
    • Conventional Imaging
    • Object Classification
    • Histogram Of Oriented Gradients
    • Spatial Resolution
    • Linear Gradient
    • Raw Images
    • Inverter
    • Extra Information
    • Supply Voltage
    • Vertical Gradient
    • Gradient Calculation
    • Gradient Computation
    • Illumination Changes
    • Charge Injection
    • Image Gradient
    • Image Compression
    • Decision Level
    • Pixel Difference
    • Image Pyramid
    • Image Pyramids
    • Horizontal Gradient
    • Linear Image
    • Gradient Components
    • Pixel Array
    • Deep Neural Network
    • Digital Control
    • Charge Redistribution
    • Capacitive Coupling
    • AC-coupled
    • Electrostatic Coupling
• authors:
  • Author Name: Christopher Young
    Affiliation: Department of Electrical Engineering, Stanford University, Stanford, CA, USA
    Author URL: https://ieeexplore.ieee.org/author/37086372228
    ID: 37086372228
    Order: 1
    Author Affiliations:
    • Department of Electrical Engineering, Stanford University, Stanford, CA, USA
  • Author Name: Alex Omid-Zohoor
    Affiliation: Department of Electrical Engineering, Stanford University, Stanford, CA, USA
    Author URL: https://ieeexplore.ieee.org/author/37086372239
    ID: 37086372239
    Order: 2
    Author Affiliations:
    • Department of Electrical Engineering, Stanford University, Stanford, CA, USA
  • Author Name: Pedram Lajevardi
    Affiliation: Research and Technology Center, Robert Bosch LLC, Sunnyvale, CA, USA
    Author URL: https://ieeexplore.ieee.org/author/38106129000
    ID: 38106129000
    Order: 3
    Author Affiliations:
    • Research and Technology Center, Robert Bosch LLC, Sunnyvale, CA, USA
  • Author Name: Boris Murmann
    Affiliation: Department of Electrical Engineering, Stanford University, Stanford, CA, USA
    Author URL: https://ieeexplore.ieee.org/author/37300093100
    ID: 37300093100
    Order: 4
    Author Affiliations:
    • Department of Electrical Engineering, Stanford University, Stanford, CA, USA

Image Sensor

• sensor_type: CMOS
• resolution: 240 x 320
• dynamic_range: 59.6 dB
• pixel_size: 5 µm
• dark_current: Not specified

Optical Data

• focal_length: Not specified
• aperture: Not specified
• field_of_view: Not specified
• distortion: Not specified

Performance Metrics

• frame_rate: 207 fps
• power_consumption: 99 pJ/pixel
• noise: 9 ENOB

Applications & Benefits

• cell_imaging: Used for low-power, always-on object detection applications that can support mobile devices through image pyramid generation.
• benefits: Significantly compressed data for comparable accuracy with reduced data movement and computation leads to a 3.3x energy reduction in backend energy.

Supporting Organizations

• supported_by: Robert Bosch LLC, Systems on Nanoscale Information Fabrics (SONIC), MARCO, DARPA.

Manuscript Details

• publication_date: Nov. 2019

Relevancy Score

• score: 10

Processed JSON Filename

request_8f9d501b-3a17-4268-8bf8-868e63dcbd4b-a_datacompressive_15275bit_loggradient_qvga_image_sensor_with_multiscale_readout_for_alwayson_object_detection.json