Browsing interface for item counterparts having different scales and lengths

Patent metadata

Field	Value
Publication number	US-10026176-B2
Application number	US-201615064490-A
Country	US
Kind code	B2
Filing date	Mar 8, 2016
Priority date	Mar 8, 2016
Publication date	Jul 17, 2018
Grant date	Jul 17, 2018

Abstract

Features are disclosed for an automatic segmentation and alignment of images for display via an interface. The images may have different scales and lengths. As such, items shown in the images, such as clothing, may not be depicted in a uniform way. Segmentation of the images into image portions where a portion of an image shows a specific item is described. The segmentation may be achieved using models and/or complex image analysis. To provide a realistic view of the subject when image segments are presented together on an interface, additional alignment of the image segments may be performed. The alignment may be achieved using models and/or complex image analysis.

First claim

What is claimed is:

1. A computer-implemented method comprising:
under control of a computing device executing specific computer-executable instructions,
receiving a first image of a human subject and a second image of the human subject, wherein the human subject includes a trunk and legs;
generating a normalized first image from the first image and a normalized second image from the second image, the normalized first image and the normalized second image showing the human subject at a common scale;
receiving an image segmentation model for segmenting the normalized first image and the normalized second image;
generating a first image portion using the image segmentation model and the normalized first image, the first image portion showing the trunk of the human subject;
generating a second image portion using the image segmentation model and the normalized second image, the second image portion showing the legs of the human subject;
causing the first image portion and the second image portion to be presented via a display; and
causing the first image portion to be presented via the display in vertical alignment with the second image portion, wherein causing the first image portion to be presented via the display in vertical alignment with the second image portion comprises:
detecting a first pose of the human subject shown in the first image;
detecting a second pose of the human subject shown in the second image;
identifying a first centerline for the first pose;
identifying a second centerline for the second pose; and
causing the first image portion and the second image portion to be presented via the display with the first centerline and the second centerline aligned.

2. The computer-implemented method of claim 1 , further comprising:
receiving first depth information for the first image and second depth information for the second image,
wherein the first depth information and the second depth information each indicate a distance of the subject from one or more sensors used to capture the first image and the second image,
wherein generating a normalized first image and a normalized second image comprises scaling at least one of the first image and the second image such that the distance, from the one or more sensors, of the subject as shown in the normalized first image is equal to the distance, from the one or more sensors, of the subject as shown in the normalized second image.

3. The computer-implemented method of claim 1 , wherein the image segmentation model comprises one of:
a skeletal model for the human subject, the skeletal model identifying a location for the trunk and legs of the subject, wherein generating the first image portion of the normalized first image includes mapping the skeletal model onto the normalized first image to identify a location of the trunk, and wherein generating the second image portion of the normalized second image includes mapping the skeletal model onto the normalized second image to identify a location of the legs;
a computer vision model, the computer vision model identifying, for a given image, the trunk and legs of the subject, wherein generating the first image portion of the normalized first image includes processing the normalized first image using the computer vision model to identify a location of the trunk, and wherein generating the second image portion of the normalized second image includes processing the normalized second image using the computer vision model to identify a location of the legs; and
a ratio model indicating a first portion of an image as a trunk area and a second portion of the image a leg area,
wherein generating the first image portion of the normalized first image includes:
identifying pixels from the first image included in the trunk area indicated by the ratio model,
generating the first image portion using the pixels from the first image included in the trunk area; and

wherein generating the second image portion of the normalized second image includes:
identifying pixels from the second image included in the leg area indicated by the ratio model, and
generating the second image portion using the pixels from the second image included in the leg area.

4. The computer-implemented method of claim 1 , wherein detecting a pose of the human subject shown in an image comprises:
receiving a pose definition identifying the pose, the pose definition including information identifying a first shoulder position relative to a second shoulder position for a subject positioned in the pose and a centerline for the pose;
identifying a first location within the image showing a first shoulder of the human subject;
identifying a second location within the image showing a second shoulder of the human subject; and
identifying the pose definition by comparing the first location and the second location with the information identifying the first shoulder position relative to the second should position.

5. The computer-implemented method of claim 1 , further comprising:
identifying a first pixel in the first image portion located within a threshold distance from a second pixel in the second image portion;
blending a color value of the first pixel with a color value of the second pixel to generate a blended color value; and
assigning the first pixel the blended color value.

6. A computer-implemented method comprising:
under control of a computing device executing specific computer-executable instructions,
generating a first image portion from a first image, the first image portion showing a first portion of a subject;
generating a second image portion from a second image, the second image portion showing a second portion of the subject;
causing presentation of the first image portion and the second image portion via a display;
determining a first centerline for the first image portion;
determining a second centerline for the second image portion; and
causing a horizontal adjustment to at least one of the first image portion or the second image portion, as presented via the display, to align the first centerline and the second centerline, wherein the first image portion as aligned with the second image portion provides a single view of the subject including the first portion of the subject and the second portion of the subject.

7. The computer-implemented method of claim 6 , further comprising:
receiving a skeletal model for the subject, the skeletal model identifying a location for a portion of the subject,
wherein generating the first image portion includes mapping the skeletal model onto the first image to identify a location of the first portion of the subject shown in the first image, and
wherein generating the second image portion includes mapping the skeletal model onto the second image to identify a location of the second portion of the subject shown in the second image.

8. The computer-implemented method of claim 7 , further comprising:
mapping the skeletal model onto the second image to identify a location of the first portion of the subject shown in the second image;
determining an adjustment factor indicating a positional adjustment to the second image to align the location of the first portion of the subject shown in the second image and the first portion of the subject shown in the first image; and
causing an adjustment to the second image portion, as presented via the display, using the adjustment factor.

9. The computer-implemented method of claim 6 , further comprising:
receiving a ratio model indicating a first area of an image including the first portion of the subject and a second area of the image including the second portion of the subject,
wherein generating the first image portion includes:
ide

Assignees

• Amazon Tech Inc

Inventors

• Davidson Alexander Adrian Hugh
• Dorner Charles Shearer
• Gradt Douglas J
• Lee Jongwoo
• Manolis Eva
• Mcdowell Maggie
• Yao Ning

Classifications

• G06T11/10

  Texturing; Colouring; Generation of textures or colours (retouching, inpainting or scratch removal G06T5/77) · CPC title

• G06T7/344 Primary

  involving models · CPC title

• G06T7/11

  Region-based segmentation · CPC title

• G06T2207/10028

  Range image; Depth image; 3D point clouds · CPC title

• G06T7/75

  involving models · CPC title