Abstract: Provided is a method of generating high definition maps, which can be used in autonomous driving. The method includes obtaining consecutive mapping data generated by a sensor installed on a vehicle at consecutive positions. The mapping data is used to generate range scan poses and GPS positions of the vehicle at the consecutive positions. The method further includes generating consecutive optimized poses of the vehicle at the consecutive positions according to the range scan poses and the GPS positions of the vehicle. A map is then generated by stitching the consecutive mapping data based on the optimized poses.

Abstract: A method of enhancing spatial resolution of an image. An image is accessed. The image comprises a set of confident pixels and an unconfident pixel positioned adjacent to each confident pixel of the set of confident pixels. The set of confident pixels comprises four or fewer pixels and corresponds to a plurality of known depth values, the unconfident pixel comprises an unknown depth value. A best depth hypothesis for the unconfident pixel is determined using a filter module in which the plurality of known depth values are used as depth hypotheses. The unconfident pixel is updated as a confident pixel using the best depth hypothesis.

Abstract: An alpha matte is generated from image forming elements of an image. For each of one or more of the image forming elements: a respective representative foreground value is determined from one or more of the image forming element values; the respective representative foreground value and the value of the image forming element are normalized with respect to a threshold level; and a respective value of the alpha matte is generated from an evaluation of the normalized image forming element value in relation to the normalized representative foreground value.

Abstract: Model-based stereo matching from a stereo pair of images of a given object, such as a human face, may result in a high quality depth map. Integrated modeling may combine coarse stereo matching of an object with details from a known 3D model of a different object to create a smooth, high quality depth map that captures the characteristics of the object. A semi-automated process may align the features of the object and the 3D model. A fusion technique may employ a stereo matching confidence measure to assist in combining the stereo results and the roughly aligned 3D model. A normal map and a light direction may be computed. In one embodiment, the normal values and light direction may be used to iteratively perform the fusion technique. A shape-from-shading technique may be employed to refine the normals implied by the fusion output depth map and to bring out fine details. The normals may be used to re-light the object from different light positions.

Abstract: Model-based stereo matching from a stereo pair of images of a given object, such as a human face, may result in a high quality depth map. Integrated modeling may combine coarse stereo matching of an object with details from a known 3D model of a different object to create a smooth, high quality depth map that captures the characteristics of the object. A semi-automated process may align the features of the object and the 3D model. A fusion technique may employ a stereo matching confidence measure to assist in combining the stereo results and the roughly aligned 3D model. A normal map and a light direction may be computed. In one embodiment, the normal values and light direction may be used to iteratively perform the fusion technique. A shape-from-shading technique may be employed to refine the normals implied by the fusion output depth map and to bring out fine details. The normals may be used to re-light the object from different light positions.

Abstract: Producing high quality depth maps. A time of flight camera and stereo cameras of a stereo camera arrangement are mapped into a common coordinate system. A correspondence between an RGB image taken from the time of flight camera and an RGB image taken from a reference stereo camera of the stereo camera arrangement is established. Intensity values of the plurality of pixels of a stereo camera left of the reference stereo camera, a stereo camera right of the reference stereo camera and the reference stereo camera are determined. These intensity values of the stereo camera left of the reference stereo camera and the stereo camera right of the reference stereo camera are compared to the reference camera. Based on the comparing, matching intensity values are determined.

Abstract: Various embodiments of the present invention related to methods for obtaining shadow invariant and shadow-shading invariant images from a single camera image taken with a narrow-band camera sensor and assuming Plankian illuminating conditions.

Abstract: A system for filtering an image using a bilateral filter includes a receiver and a bilateral filter. The receiver is for receiving an image having a plurality of pixels. The bilateral filter is run in constant time and comprises a pixel value range determinor and a spatial filter. The pixel value range determinor is configured to determine a range of pixel values of the plurality of pixels. The spatial filter is configured to spatially filter at least one of the plurality of pixels. The spatial filter is further configured to perform over a plurality of pixel values in the range of pixel values.

Abstract: An alpha matte is generated from image forming elements of an image. For each of one or more of the image forming elements: a respective representative foreground value is determined from one or more of the image forming element values; the respective representative foreground value and the value of the image forming element are normalized with respect to a threshold level; and a respective value of the alpha matte is generated from an evaluation of the normalized image forming element value in relation to the normalized representative foreground value.

Abstract: Various embodiments of the present invention related to methods for obtaining shadow invariant and shadow-shading invariant images from a single camera image taken with a narrow-band camera sensor and assuming Plankian illuminating conditions.