Browse Prior Art Database

A method for interactive clothes image search assisted by virtual 3D human shape model Disclosure Number: IPCOM000238317D
Publication Date: 2014-Aug-18
Document File: 7 page(s) / 150K

Publishing Venue

The Prior Art Database


CBIR (Content based Image Retrieval) becomes popular in recent years as there is a strong visual search requirement from e-commerce and mobile applications. Content based image search is still a quite challenging problem due to the gap between image pixels and semantic meaning of the content in the image. Usually, to make CBIR technology more accurate, object segmentation by automatic algorithm or human-machine interaction is needed before image content matching. Automatic visual object segmentation encounters huge challenges in real applications due to cluttered background, low contrast imaging condition, and so on. Therefore, human-machine interaction based object segmentation is still a practical approach in real use. This disclosure reveals an interactive human-object segmentation method for clothing search in the context of mobile e-commerce.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 100% of the total text.

Page 01 of 7

A method for interactive clothes image search assisted by virtual

A method for interactive clothes image search assisted by virtual

In this method, a deformable 3D virtual human shape with several dressing styles (e.g. one-piece dress, coat-trousers, coat-skirt, and so on) is provided to the user. The user can chose the right dressing style for the photo he/she wants to do accurate clothing search, and then he/she can rotate, scale and translate the virtual 3D human shape by touch screen of mobile device to best fit the human object in the photo. After the interaction process, 3D human shape can be projected to the 2D image plane and parts of interests (such as coat, trousers, shoes, etc) in the photo can be segmented according to the prior knowledge of the 3D human shape, then the image regions of these parts of interests can be sent to the image search engine for accurate content retrieval.

333D human shape model D human shape model


Page 02 of 7


Page 03 of 7


Page 04 of 7


Page 05 of 7


Page 06 of 7


Page 07 of 7