Depth from motion for augmented reality for handheld user devices
A handheld user device includes a monocular camera to capture a feed of images of a local scene and a processor to select, from the feed, a keyframe and perform, for a first image from the feed, stereo matching using the first image, the keyframe, and a relative pose based on a pose associated with the first image and a pose associated with the keyframe to generate a sparse disparity map representing disparities between the first image and the keyframe. The processor further is to determine a dense depth map from the disparity map using a bilateral solver algorithm, and process a viewfinder image generated from a second image of the feed with occlusion rendering based on the depth map to incorporate one or more virtual objects into the viewfinder image to generate an AR viewfinder image. Further, the processor is to provide the AR viewfinder image for display.
1. A method for providing an augmented reality (AR) experience at a handheld user device, the method comprising:
capturing, via the handheld user device, a feed of images of a local scene;
selecting, from the feed, a keyframe;
performing, for a first image from the feed of images, stereo matching using the first image, the keyframe, and a relative pose based on a pose associated with the first image and a pose associated with the keyframe to generate a sparse disparity map representing disparities between the first image and the keyframe;
determining a dense depth map from the disparity map using a bilateral solver algorithm;
processing a viewfinder image generated from a second image of the feed with occlusion rendering based on the depth map to incorporate one or more virtual objects into the viewfinder image to generate an AR viewfinder image; and
displaying, at the handheld user device, the AR viewfinder image.
2. The method of claim 1 , further comprising:
polar rectifying the keyframe and the first image; and
wherein performing stereo matching comprising performing stereo matching using polar rectified representations of the keyframe and the first image.
3. The method of either of claim 1 , wherein determining the depth map comprises:
generating a sparse depth map from the disparity map using triangulation;
applying the bilateral solver algorithm to the sparse depth map to generate a bilateral grid of depths; and
slicing the bilateral grid of depths with the second image to generate the depth map.
4. The method of claim 3 , wherein the bilateral solver algorithm comprises a planar bilateral solver algorithm that is based on plane-fitting each pixel in the sparse depth map.
5. The method of claim 1 , wherein selecting the keyframe comprises:
selecting the keyframe based on minimization of a cost function that implements at least one of: a baseline distance between two candidate frames, a time difference between the capture of two candidate frames, a fractional overlap of image areas of two candidate frames based on viewing frustums, and a measured error of pose-tracking statistics for two candidate frames.
6. The method of claim 1 , wherein selecting the keyframe, performing stereo matching, determining the depth map, and processing the viewfinder image are performed in real-time by a central processing unit (CPU) of the handheld user device.
7. A handheld user device, comprising:
a monocular camera to capture a feed of images of a local scene;
a display panel;
a memory to store a software application; and
a processor coupled to the memory and to the monocular camera, wherein the processor is to execute instructions of the software application to:
select, from the feed, a keyframe;
perform, for a first image from the feed of images, stereo matching using the first image, the keyframe, and a relative pose based on a pose associated with the first image and a pose associated with the keyframe to generate a sparse disparity map representing disparities between the first image and the keyframe;
determine a dense depth map from the disparity map using a bilateral solver algorithm;
process a viewfinder image generated from a second image of the feed with occlusion rendering based on the depth map to incorporate one or more virtual objects into the viewfinder image to generate an AR viewfinder image; and
provide the AR viewfinder image for display at the display panel.
8. The handheld user device of claim 7 , wherein the processor is to execute instructions of the software application further to:
polar rectify the keyframe and the first image; and
the stereo matching uses polar rectified representations of the keyframe and the first image.
9. The handheld user device of claim 7 , wherein the processor is to determine the depth map by:
generating a sparse depth map from the disparity map using triangulation;
applying the bilateral solver algorithm to the sparse depth map to generate a bilateral grid of depths; and
slicing the bilateral grid of depths with the second image to generate the depth map.
10. The handheld user device of claim 9 , wherein the bilateral solver algorithm comprises a planar bilateral solver algorithm that is based on plane-fitting each pixel in the sparse depth map.
11. The handheld user device of claim 7 , wherein the processor is to select the keyframe based on minimization of a cost function that implements at least one of: a baseline distance between two candidate frames; a time difference between the capture of two candidate frames; a fractional overlap of image areas of two candidate frames based on viewing frustums; and a measured error of pose-tracking statistics for two candidate frames.
12. The handheld user device of claim 7 , wherein the processor is a central processing unit (CPU).
13. The handheld user device of claim 12 , wherein the CPU is to determine the depth map and process the viewfinder image in real-time.
14. The handheld user device of claim 7 , wherein the handheld user device is one of a compute-enabled cellular phone, a tablet computer, and a portable gaming device.
15. A non-transitory computer-readable storage medium storing a set of executable instructions, the set of executable instructions configured to manipulate a processor of a handheld user device to:
select a keyframe from a feed of images captured by a monocular camera;
perform, for a first image from the feed of images, stereo matching using the first image, the keyframe, and a relative pose based on a pose associated with the first image and a pose associated with the keyframe to generate a sparse disparity map representing disparities between the first image and the keyframe;
determine a dense depth map from the disparity map using a bilateral solver algorithm;
process a viewfinder image generated from a second image of the feed with occlusion rendering based on the depth map to incorporate one or more virtual objects into the viewfinder image to generate an AR viewfinder image; and
provide the AR viewfinder image for display at a display panel of the handheld user device.
16. The non-transitory computer-readable storage medium of claim 15 , wherein the executable instructions are configured to manipulate the processor further to:
polar rectify the keyframe and the first image; and
the stereo matching uses polar rectified representations of the keyframe and the first image.
17. The non-transitory computer-readable storage medium of claim 15 , wherein the executable instructions are to manipulate the processor to determine the depth map by:
generating a sparse depth map from the disparity map using triangulation;
applying the bilateral solver algorithm to the sparse depth map to generate a bilateral grid of depths; and
slicing the bilateral grid of depths with the second image to generate the depth map.
18. The non-transitory computer-readable storage medium of claim 17 , wherein the bilateral solver algorithm comprises a planar bilateral solver algorithm that is based on plane-fitting each pixel in the sparse depth map.
19. The non-transitory computer-readable storage medium of claim 15 , wherein the processor is to select the keyframe based on minimization of a cost function that implements at least one of: a baseline distance between two candidate frames; a time difference between the capture of two candidate frames; a fractional overlap of image areas of two candidate frames based on viewing frustums; and a measured error of pose-tracking statistics for two candidate frames.
20. The non-transitory computer-readable storage medium of claim 15 , wherein the processor is a central processing unit (CPU).