Three dimensional CMOS image processor for feature detection

• US 8,942,481 B2
• Filed: 03/11/2012
• Issued: 01/27/2015
• Est. Priority Date: 03/11/2012
• Status: Expired due to Fees

• Current Assignee: Consejo Superior de Investigaciones Científicas,, Universidad De Santiago De Compostela (Tecnocontrol Servicios SA), Universidad de Sevilla
• Original Assignee: Consejo Superior de Investigaciones Científicas,, Universidad De Santiago De Compostela (Tecnocontrol Servicios SA), Universidad de Sevilla
• Inventors: Surez Cambre, Manuel, Brea Snchez, Vctor Manuel, Pardo Seco, Fernando Rafael, Rodrguez Vzquez, ngel, Carmona Galn, Ricardo
• Primary Examiner(s): Bella, Matthew
• Assistant Examiner(s): HEIDEMANN, JASON E
• Application Number:
  US13/417,279
• Publication Number:
  US 20130236048A1
• Time in Patent Office: 1,052 Days
• Field of Search: None
• US Class Current: 382/190
• CPC Class Codes: G06V 10/462 Salient features, e.g. scale invariant feature transforms [SIFT]
• CPC Class Codes: G06V 10/955 using specific electronic processors

Abstract

Disclose embodiments include an image processor for feature detection comprising a single non-planar chip containing a plurality of integrated sensing and processing resources across two or more layers adapted to capture image frames and extract image features. In a particular embodiment, the non-planar chip is a three dimensional CMOS integrated circuit (3D CMOS IC) with vertical distribution of sensing and processing resources across two or more vertical integrated circuit layers. The 3D CMOS IC implements two or more feature detectors in a single chip by reusing a plurality of circuits employed for gradient and keypoint detection. Feature detectors include a scale invariant feature transform detector (SIFT), a Harris-based feature detector, and a Hessian-based feature detector.

Claims

1. An image processor for feature detection comprising a single non-planar chip containing a plurality of integrated sensing and processing resources across two or more layers adapted to capture image frames and extract image features, wherein said non-planar chip is a three dimensional CMOS integrated circuit with vertical distribution of sensing and processing resources across two or more vertical integrated circuit layers and wherein said three dimensional CMOS integrated circuit comprises an array of in-pixel processors and a plurality of circuits that in parallel realize a scale invariant feature transform detector (SIFT), a Harris-based feature detector, and a Hessian-based feature detector.

2. The image processor of claim 1, wherein said three dimensional CMOS integrated circuit implements said feature detectors by reusing a plurality of circuits employed for gradient and keypoint detection.

3. The image processor of claim 1, wherein said three dimensional CMOS integrated circuit implements a pixel-per-processor architecture to achieve massive pixel-level parallelism and pixel-level operation capability.

4. The image processor of claim 1, wherein said plurality of circuits include a multiple purpose switched-capacitor network.

5. The image processor of claim 4, wherein said multiple purpose switched-capacitor network implements Gaussian pyramid generation, in-pixel correlated double sampling (CDS), and a pixel-level analog memory.

6. The image processor of claim 5, wherein said multiple purpose switched-capacitor network has a double Forward-Euler configuration.

7. The image processor of claim 6, wherein said three dimensional CMOS integrated circuit includes shared reused circuits for in-pixel analog-to-digital conversion and frame buffering.

8. The image processor of claim 7, wherein said circuits of said three dimensional CMOS integrated circuit are arranged in three layers with (a) a tier for acquisition, correlated double sampling, Gaussian filtering, and analog-to-digital conversion;

(b) a tier for digital frame buffering, keypoint detection, Difference-of-Gaussians (DoGs) calculation, gradient calculation, and shared analog-to-digital conversion; and

(c) two tiers with a DRAM memory.

9. A method for achieving efficient image feature detection on an integrated circuit, comprising:

(a) implementing a single non-planar chip containing a plurality of integrated sensing and processing resources across two or more layers adapted to capture image frames and extract image features, wherein said non-planar chip is a three dimensional CMOS integrated circuit with vertical distribution of sensing and processing resources across two or more vertical integrated circuit layers and wherein said three dimensional CMOS integrated circuit comprises an array of in-pixel processors and a plurality of circuits that in parallel realize a scale invariant feature transform detector (SIFT), a Harris-based feature detector, and a Hessian-based feature detector; and

(b) reusing a plurality of circuits in said non-planar chip to achieve concurrent implementation of gradient and keypoint detection; and

correlated double sampling (CDS) and analog-to-digital conversion.

10. The method of claim 9, wherein said three dimensional CMOS integrated circuit implements a pixel-per-processor to achieve massive pixel-level parallelism and pixel-level operation capability;

said plurality of circuits include a multiple purpose switched-capacitor network;

said multiple purpose switched-capacitor network implements Gaussian pyramid generation, in-pixel correlated double sampling (CDS), and a pixel-level analog memory;

said three dimensional CMOS integrated circuit includes shared circuits for in-pixel analog to digital conversion and frame buffering;

said multiple purpose switched-capacitor network implements Gaussian pyramid generation, in-pixel correlated double sampling (CDS), and a pixel-level analog memory;

said multiple purpose switched-capacitor network has a double Forward-Euler configuration;

said three dimensional CMOS integrated circuit includes shared circuits for in-pixel analog to digital conversion and frame buffering; and

said circuits of said three dimensional CMOS integrated circuit are arranged in three layers with (a) a tier for acquisition, correlated double sampling, Gaussian filtering, and analog-to-digital conversion;

(b) a tier for digital frame buffering, keypoint detection, Difference-of-Gaussians (DoGs) calculation, gradient calculation, and shared analog-to-digital conversion; and

(c) a tier with a DRAM memory.