Machine-learned depth dealiasing

A computer system configured to de-alias depth ambiguities included within a phase image, the computer system comprising:

• one or more processors;
• and
• one or more computer-readable hardware storage devices that store instructions that are executable by the one or more processors to cause the computer system to at least:
• access a phase image comprising phase differences that represent changes in depth within overlapping sinusoidal periods occurring between a reference light and a reflected light, the phase image including ambiguities that occur as a result of multiple different depths represented by the phase image sharing a same phase difference value;
• cause a machine learning component to de-alias the ambiguities by determining de-aliasing intervals for different sinusoidal periods included in the phase image;
• and
• generate a depth map based on the phase image and any resulting de-aliasing intervals generated by the machine learning component.

A method for de-aliasing depth ambiguities included within infrared depth images, the method comprising:

• accessing a phase image that was generated based on detected phase differences, the phase differences representing changes in depth within overlapping sinusoidal periods between a reference light and a reflected light, the phase image including ambiguities as a result of multiple different depths represented by the phase image sharing a same phase difference value;
• causing a machine learning component to de-alias the ambiguities in the phase image by determining, for each pixel in the phase image, a corresponding de-aliasing interval;
• and
• generating a depth map based on the phase image and any resulting de-aliasing intervals generated by the machine learning component.

A computer system configured to de-alias depth ambiguities included within a phase image, the computer system comprising:

• one or more processors;
• and
• one or more computer-readable hardware storage devices that store instructions that are executable by the one or more processors to cause the computer system to at least:
• access a phase image that was generated based on detected phase differences, the phase differences representing changes in depth within overlapping sinusoidal periods between a reference light and a reflected light, the phase image including ambiguities as a result of multiple different depths represented by the phase image sharing a same phase difference value;
• feed the phase image as input to a machine learning component configured to de-alias the ambiguities by determining, for each pixel in the phase image, a corresponding de-aliasing interval;
• and
• generate a depth map based on the phase image and any resulting de-aliasing intervals generated by the machine learning component.