A kind of picture browsing method

Abstract

The invention discloses a picture browsing method, comprising: detecting a human face in a target picture set based on a human face feature vector; selecting a specified human face from the target picture set after the detected human face, and obtaining the human face feature vector of the specified human face ; Calculate the similarity between the specified face and other faces detected in the target picture set based on the face feature vector; sort and display the pictures in the target picture set based on the similarity. The invention gets rid of the complexity of manual labeling and avoids the high error rate of blind automatic grouping, and realizes picture browsing based on the similarity of human faces.

Description

technical field [0001] The invention relates to the technical field of authentication and identification, in particular to a picture browsing method. Background technique [0002] With the popularization of digital photography equipment and the development of memory, more and more pictures are taken and stored by users, and pictures of people account for a relatively large proportion of them. When a user is viewing a picture, digital devices such as a mobile phone, a computer, and an electronic photo frame usually present the picture according to the shooting or saving time for the user to browse. But users are often more interested in the characters in the picture. Therefore, based on face recognition technology, the browsing method of clustering multiple pictures of the same person and then sorting them in descending order of similarity can better meet the potential needs of users. [0003] Some smart phone photo albums currently on the market support image browsing base...

AI Technical Summary

Problems solved by technology

[0004] In the above method of realizing picture browsing through character tags, the main disadvantage of the first method is that it requires a huge labor cost in the early stage. Users need to mark each picture in order to have a better tag grouping effect, which consumes time and affects User experience; the second method is mainly restricted by the performance of face detection and grouping. Most of the products on the market have missed face detection (that is, not detecting all faces in the picture), false detection (not detecting all faces in the image) Detection and judgment as human faces), grouping errors (different people are classified into the same group or the same person is divided into different groups), etc.

Moreover, the focus of the two methods is on group viewing. The similarity calculation is only used to obtain the group, and the pictures cannot be sorted by similarity.

At the same time, those pictures without tags may be ignored when browsing

Moreover, in the second method, for some people who are not interested (such as passers-by appearing in the photo) or false detection (half face or not face) will lead to inaccurate picture grouping results and affect user experience

For the processing of multi-person photos or group photos, group browsing of pictures based on tags is also likely to cause trouble for users: some users may only want to group individual photos of a certain person into a group, and the smart album also inserts group photos into this group; And some users may want to add both group photos and single photos of a person to the same group

Images