Efficient SIMD implementation of 3X3 non maxima suppression of sparse 2D image feature points

• US 10,521,688 B2
• Filed: 05/25/2018
• Issued: 12/31/2019
• Est. Priority Date: 04/19/2016
• Status: Active Grant

• Current Assignee: Texas Instruments, Inc.
• Original Assignee: Texas Instruments, Inc.
• Inventors: Poddar, Deepak Kumar, Swami, Pramod Kumar, Viswanath, Prashanth
• Primary Examiner(s): Carter, Aaron W
• Application Number:
  US15/989,551
• Publication Number:
  US 20190005349A1
• Time in Patent Office: 585 Days
• 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

This invention transforms a list of feature points in raster scan order into a list of maxima suppressed feature points. A working buffer has two more entries than the width of the original image. Each entry is assigned to an x coordinate of the original image. Each entry stores a combined y coordinate and reliability score for each feature point in the original list. This process involves a forward scan and a backward scan. For each original feature point its x coordinate defines the location within the working buffer where neighbor feature points would be stored if they exist. The working buffer initial data and the y coordinates assure a non-suppress comparison result if the potential neighbors are not actual neighbors. For actual neighbor data, the y coordinates match and the comparison result depends solely upon the relative reliability scores.

Claims

1. A method comprising:

receiving a list of feature points of an image, wherein each feature point corresponds to a pixel of the image and has properties including an x-coordinate, a y-coordinate, and a reliability score;

initializing a valid list, the initialized valid list indicating that all of the feature points of the list are non-suppressed feature points;

performing a first scan of the list of feature points in a first direction in which, for each feature point, at least one property of the feature point is compared a corresponding property of each of a first plurality of feature points, wherein each of the first plurality of feature points correspond to a respective one of a first plurality of pixels, and wherein each of the first plurality of pixels neighbor the pixel to which the feature point corresponds;

updating the valid list in response to the first scan to obtain an intermediate valid list;

performing a second scan of the list of feature points in a second direction in which, for each feature point, at least one property of the feature point is compared a corresponding property of each of a second plurality of feature points, wherein each of the second plurality of feature points correspond to a respective one of a second plurality of pixels, and wherein each of the second plurality of pixels neighbor the pixel to which the feature point corresponds;

updating the intermediate valid list in response to the second scan to obtain a final valid list, wherein the final valid list indicates that at least one of the feature points of the list is to be suppressed; and

generating a non-maxima suppressed feature point list that includes only feature points from the list that the final valid list indicates are not to be suppressed.

2. The method of claim 1, wherein the at least one property of the feature point is the y-coordinate and the reliability score.

3. The method of claim 1, wherein the at least one property of the feature point is a combined y-coordinate and the reliability score.

4. The method of claim 1, wherein the first direction is a forward direction and the second direction is a backward direction.

5. The method of claim 1, wherein each of the pixels of the first plurality of pixels are different from each of the pixels of the second plurality of pixels.

6. The method of claim 1, wherein the sum of the first plurality of pixels and the second plurality of pixels is 8.

7. The method of claim 6, wherein the first plurality of pixels includes four pixels and the second plurality of pixels includes four pixels, with each pixel of the first plurality of pixels being different from each pixel of the second plurality of pixels.

8. The method of claim 1, wherein:

the first plurality of pixels includes a first pixel that is a left neighbor pixel with respect to the pixel to which the feature point corresponds, a second pixel that is a top left neighbor pixel with respect to the pixel to which the feature point corresponds, a third pixel that is a top neighbor pixel with respect to the pixel to which the feature point corresponds, and a fourth pixel that is a top right neighbor pixel to the pixel to which the feature point corresponds; and

the second plurality of pixels includes a fifth pixel that is a right neighbor pixel with respect to the pixel to which the feature point corresponds, a sixth pixel that is a bottom right neighbor pixel with respect to the pixel to which the feature point corresponds, a seventh pixel that is a bottom neighbor pixel with respect to the pixel to which the feature point corresponds, and an eighth pixel that is a bottom left neighbor pixel to the pixel to which the feature point corresponds.

9. An electronic device comprising:

a processor;

a memory to store image data and program instructions that, when executed by the processor, cause the processor to;

retreive a list of feature points of an image from the image data stored in the memory, wherein each feature point corresponds to a pixel of the image and has properties including an x-coordinate, a y-coordinate, and a reliability score;

initialize a valid list, the initialized valid list indicating that all of the feature points of the list are non-suppressed feature points;

perform a first scan of the list of feature points in a first direction in which, for each feature point, at least one property of the feature point is compared a corresponding property of each of a first plurality of feature points, wherein each of the first plurality of feature points correspond to a respective one of a first plurality of pixels, and wherein each of the first plurality of pixels neighbor the pixel to which the feature point corresponds;

update the valid list in response to the first scan to obtain an intermediate valid list;

perform a second scan of the list of feature points in a second direction in which, for each feature point, at least one property of the feature point is compared a corresponding property of each of a second plurality of feature points, wherein each of the second plurality of feature points correspond to a respective one of a second plurality of pixels, and wherein each of the second plurality of pixels neighbor the pixel to which the feature point corresponds;

update the intermediate valid list in response to the second scan to obtain a final valid list, wherein the final valid list indicates that at least one of the feature points of the list is to be suppressed; and

generate a non-maxima suppressed feature point list that includes only feature points from the list that the final valid list indicates are not to be suppressed.

10. The electronic device of claim 9, wherein the at least one property of the feature point is the y-coordinate and the reliability score.

11. The electronic device of claim 9, wherein the at least one property of the feature point is a combined y-coordinate and the reliability score.

12. The electronic device of claim 9, wherein the first direction is a forward direction and the second direction is a backward direction.

13. The electronic device of claim 9, wherein each of the pixels of the first plurality of pixels are different from each of the pixels of the second plurality of pixels.

14. The electronic device of claim 9, wherein the sum of the first plurality of pixels and the second plurality of pixels is 8.

15. The electronic device of claim 14, wherein the first plurality of pixels includes four pixels and the second plurality of pixels includes four pixels, with each pixel of the first plurality of pixels being different from each pixel of the second plurality of pixels.

16. The electronic device of claim 9, wherein:

the first plurality of pixels includes a first pixel that is a left neighbor pixel with respect to the pixel to which the feature point corresponds, a second pixel that is a top left neighbor pixel with respect to the pixel to which the feature point corresponds, a third pixel that is a top neighbor pixel with respect to the pixel to which the feature point corresponds, and a fourth pixel that is a top right neighbor pixel to the pixel to which the feature point corresponds; and

the second plurality of pixels includes a fifth pixel that is a right neighbor pixel with respect to the pixel to which the feature point corresponds, a sixth pixel that is a bottom right neighbor pixel with respect to the pixel to which the feature point corresponds, a seventh pixel that is a bottom neighbor pixel with respect to the pixel to which the feature point corresponds, and an eighth pixel that is a bottom left neighbor pixel to the pixel to which the feature point corresponds.