Method and apparatus for extracting image feature

• US 9,760,795 B2
• Filed: 04/29/2014
• Issued: 09/12/2017
• Est. Priority Date: 02/24/2014
• Status: Active Grant

• Current Assignee: Electronics and Telecommunications Research Institute
• Original Assignee: Electronics and Telecommunications Research Institute
• Inventors: Lee, Seungjae, Na, Sangil, Lee, Keun Dong, Je, Sungkwan, Jung, Da-Un, Oh, Weon Geun, Suh, Young Ho, Son, Wookho
• Primary Examiner(s): Abdi, Amara
• Application Number:
  US14/265,283
• Publication Number:
  US 20150242703A1
• Time in Patent Office: 1,232 Days
• CPC Class Codes: G06F 16/583 using metadata automatically derived from the content
• CPC Class Codes: G06V 10/462 Salient features, e.g. scale invariant feature transforms [SIFT]

Abstract

A method for extracting features from an image for use in a computing device, the method comprising: producing Gaussian Scale Space (GSS) images in the type of a pyramid from the image inputted to the computing device; performing a Scale Normalized Laplacian Filtering on the GSS images; detecting interest points from the images that are subject to the Scale Normalized Laplacian Filtering; and extracting features of the image using the detected interest points.

Claims

1. A method for extracting features from an image for use in a computing device, the method comprising:

producing Gaussian Scale Space (GSS) images in the type of a pyramid from the image inputted to the computing device;

performing a Scale Normalized Laplacian Filtering on the GSS images;

detecting interest points from the images that are subject to the Scale Normalized Laplacian Filtering; and

extracting features of the image using the detected interest points,

wherein the detecting interest points comprises processing three adjacent GSS images using one storage area for GSS images, one storage area for SCLoG (Scale Normalized Laplacian of Gaussian) images, one storage area for gradients, and a plurality of temporary storage areas, and

wherein the processing three adjacent GSS images includes copying and calculating the GSS images and the SCLoG images into the temporary storage areas.

2. The method of claim 1, wherein said detecting interest points comprises:

storing adjacent first, second and third GSS images among the GSS images in a first storage space;

copying the second GSS image and the third GSS image into a second storage space;

producing first, second and third SCLOG (Scale Normalized Laplacian of Gaussian) images corresponding to the first, second and third GSS images;

storing the second SCLoG image among the SCLoG images in the second storage space; and

calculating a first gradient from the second GSS image in the second storage space to store in the first storage space.

3. The method of claim 2, wherein the first gradient comprises a magnitude gradient and an orientation gradient calculated from the second GSS image.

4. The method of claim 2, wherein said detecting interest points further comprises:

storing the second GSS image and the third GSS image in the first storage space;

producing a fourth GSS image from the third GSS image to store in the second storage space;

producing a fourth SCLog image from the fourth GSS image to store in the first storage space;

calculating a second gradient from the three GSS image in the second storage space to store in the first storage space.

5. The method of claim 4, wherein the second gradient comprises a magnitude gradient and an orientation gradient calculated from the third GSS image.

6. The method of claim 4, wherein said producing a fourth GSS image from the third GSS image comprises applying a Gaussian Smoothing to the third GSS image.

7. An apparatus for extracting features from an image for use in a computing device, the apparatus comprising:

means for producing Gaussian Scale Space (GSS) images in the type of a pyramid from the image inputted to the computing device;

means for performing a Scale Normalized Laplacian Filtering on the GSS images;

means for detecting interest points from the images that are subjected to the Scale Normalized Laplacian Filtering; and

means for extracting features of the image using the detected interest points,

wherein the means for detecting interest points is configured to process three adjacent GSS images using one storage area for GSS images, one storage area for SCLoG (Scale Normalized Laplacian of Gaussian) images, one storage area for gradients, and a plurality of temporary storage areas, by copying and calculating the GSS images and the SCLoG images into the temporary storage areas.

8. The apparatus of claim 7, wherein the means for detecting interest points is further configured to:

store adjacent first, second and third GSS images among the GSS images in a first storage space;

copy the second GSS image and the third GSS image into a second storage space;

produce first, second and third SCLOG (Scale Normalized Laplacian of Gaussian) images corresponding to the first, second and third GSS images;

store the second SCLoG image among the SCLoG images in the second storage space; and

calculate a first gradient from the second GSS in the second storage space to store in the first storage space.

9. The apparatus of claim 8, wherein the first gradient comprises a magnitude gradient and an orientation gradient calculated from the second GSS image.

10. The apparatus of claim 8, wherein the means for detecting interest points is further configured to:

store the second GSS image and the third GSS image in the first storage space;

produce a fourth GSS image from the third GSS image to store in the second storage space;

produce a fourth SCLog image from the fourth GSS image to store the first storage space;

calculate a second gradient from the three GSS image in the second storage space to store in the first storage space.

11. The apparatus of claim 10, wherein the second gradient comprises a magnitude gradient and an orientation gradient calculated from the third GSS image.

12. The apparatus of claim 10, wherein the fourth GSS image is produced by applying a Gaussian Smoothing to the third GSS image.

13. A method for extracting features from an image for use in a computing device, the method comprising:

producing Gaussian Scale Space (GSS) images in the type of a pyramid from the image inputted to the computing device;

performing a Scale Normalized Laplacian Filtering on the GSS images;

detecting interest points from the images that are subject to the Scale Normalized Laplacian Filtering; and

extracting features of the image using the detected interest points,

wherein said detecting interest points comprises;

storing adjacent first, second and third GSS images among the GSS images in a first storage space;

copying the second GSS image and the third GSS image into a second storage space;

producing first, second and third SCLOG (Scale Normalized Laplacian of Gaussian) images corresponding to the first, second and third GSS images;

storing the second SCLoG image among the SCLoG images in the second storage space; and

calculating a first gradient from the second GSS image in the second storage space to store in the first storage space.