Method for determining depth maps from stereo images with improved depth resolution in a range

Patent metadata

Field	Value
Publication number	US-10223802-B2
Application number	US-201715475847-A
Country	US
Kind code	B2
Filing date	Mar 31, 2017
Priority date	Jan 14, 2013
Publication date	Mar 5, 2019
Grant date	Mar 5, 2019

Abstract

In a method for determining a depth map from stereo images, the disparity for a pixel is selected from a predefined quantity or set of predefined discrete disparity values that are distributed over the entire predefined disparity value range, whereby the distribution is non-uniform or has at least two different distances or intervals between different adjacent disparity values. This method makes it possible to more precisely determine (with finer resolution) especially only those disparities for which a more precise determination is required.

First claim

The invention claimed is:

1. A method of determining a depth map from stereo image pairs, wherein a respective disparity for at least one respective pixel of one of the stereo image pairs is determined from a discrete quantity of predefined disparity values that are distributed over an entire predefined disparity value range with a distribution that has at least two different intervals between different adjacent ones of the disparity values.

2. The method according to claim 1 , wherein at least one of the intervals has a length of less than one pixel spacing.

3. The method according to claim 1 , wherein comparison/similarity costs for the disparity values are determined by a suitable comparison operator.

4. The method according to claim 1 , wherein the disparity values include integer disparity values and sub-pixel disparity values, comparison/similarity costs of the integer disparity values are determined by an appropriate comparison operator from one of the stereo image pairs, and comparison/similarity costs of the sub-pixel disparity values are determined by interpolation from the comparison/similarity costs of adjacent ones of the integer disparity values.

5. The method according to claim 1 , wherein effects caused by over-sampling or sub-sampling are taken into account.

6. The method according to claim 1 , wherein the disparity for the respective pixel is determined by a global or semi-global matching or correspondence method.

7. The method according to claim 6 , wherein the disparity is further refined sub-pixel precisely based on the disparity value with minimal cost among comparison/similarity costs of adjacent ones of the disparity values.

8. An apparatus for determining a depth map from stereo image pairs according to the method of claim 1 , comprising a stereo camera configured and arranged to record the stereo image pairs, and a determination unit configured and arranged to evaluate the stereo image pairs recorded by the stereo camera according to the method.

9. A method of determining a depth map from at least one stereo image pair, wherein disparities are determined from the at least one stereo image pair by selecting the disparities from among predetermined potential disparity values that are non-uniformly distributed over an entire disparity value range of the disparity values.

10. The method according to claim 9 , wherein intervals between two adjacent ones of the disparity values, in which disparities for a close range are determined, are larger than other intervals between two other adjacent ones of the disparity values in which disparities for other than the close range are determined.

11. The method according to claim 10 , wherein the close range is a maximum of 5 meters away from a stereo camera, with which the at least one stereo image pair is recorded.

12. The method according to claim 9 , wherein the disparities are determined by a global or semi-global matching or correspondence method.

13. The method according to claim 9 , wherein at least one interval between two adjacent ones of the disparity values is smaller than a pixel.

14. The method according to claim 9 , wherein effects caused by over-sampling or sub-sampling are taken into account.

15. The method according to claim 9 , wherein similarity costs for the disparities are determined.

16. The method according to claim 15 , wherein the similarity costs of one of the disparities are determined from the similarity costs of adjacent ones of the disparities.

17. The method according to claim 15 , wherein the similarity costs of one of the disparities are determined directly from the stereo image pair.

18. The method according to claim 17 , wherein the disparities and the similarity costs thereof are determined only for intervals between two adjacent disparity values, which correspond to one or more whole pixels.

19. A method of determining a depth map, comprising the steps:
a) with a stereo camera, providing at least one stereo image pair;
b) providing, in an evaluation unit, a predefined set of discrete disparity values that span a disparity value range, wherein intervals between successive ones of the disparity values include first intervals and second intervals, wherein the first intervals have a magnitude smaller than that of the second intervals;
c) in the evaluation unit, determining a respective disparity for a respective pixel of a reference image of the at least one stereo image pair, comprising selecting the respective disparity from among the discrete disparity values of the predefined set; and
d) in the evaluation unit, determining a respective depth value for the respective pixel by calculation from the respective disparity that was determined for the respective pixel.

20. The method according to claim 19 , wherein the intervals further include third intervals that have a magnitude between those of the first intervals and the second intervals.

21. The method according to claim 20 , wherein the first intervals are provided between successive ones of the disparity values corresponding to a far-range of depth, the second intervals are provided between successive ones of the disparity values corresponding to a near-range of depth, and the third intervals are provided between successive ones of the disparity values corresponding to a mid-range of depth.

22. The method according to claim 19 , wherein the first intervals are provided between successive ones of the disparity values corresponding to a far-range of depth, and the second intervals are provided between successive ones of the disparity values corresponding to a near-range of depth.

23. The method according to claim 19 , wherein the first intervals are provided in a first value range of the disparity values, and the second intervals are provided in a second value range of the disparity values different from the first value range.

24. The method according to claim 19 , wherein the first intervals are provided between successive ones of the disparity values corresponding to a first depth range in which an accuracy, a precision or a resolution of the determination of the depth value in the step d) is to be increased, and the second intervals are provided between successive ones of the disparity values corresponding to a second depth range other than the first depth range.

25. The method according to claim 19 , further comprising performing the step c) of determining the respective disparity respectively for each pixel of the reference image using the predefined set of discrete disparity values again respectively for each pixel, which comprises selecting the respective disparity for each pixel respectively from among the discrete disparity values of B the predefined set.

26. The method according to claim 25 , wherein the at least one stereo image pair includes a succession of plural stereo image pairs, and the step c) comprises performing the determining of the respective disparity respectively for each pixel of all of the stereo image pairs using the predefined set of discrete disparity values again respectively for each pixel of all of the stereo image pairs.

27. The method according to claim 25 ,
wherein the step c) further comprises forming a cost volume representing a respective cost value associated respectively with each discrete disparity value of the predefined set assigned to each pixel of the reference image, and
wherein the selecting in the

Assignees

• Conti Temic Microelectronic Gmbh

Inventors

• Lange Frederik
• Stueck Robert
• Exner Christian
• Bachmann Alexander
• Baratoff Gregory

Classifications

• G06T7/593 Primary

  from stereo images · CPC title

• G06T7/55 Primary

  from multiple images · CPC title

• G06F18/22

  Matching criteria, e.g. proximity measures · CPC title

• G06K9/52

  Physics · mapped topic

• H04N13/271

  wherein the generated image signals comprise depth maps or disparity maps · CPC title