Browse Prior Art Database

Generating a Quality Score for a Photograph Based on Learned Preferences at the Time the Photograph is Taken

IP.com Disclosure Number: IPCOM000246560D
Publication Date: 2016-Jun-17
Document File: 2 page(s) / 37K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a system that uses machine-learning techniques to establish a base set of quality criteria for photographs.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 01 of 2

Generating a Quality Score for a Photograph Based on Learned Preferences at the Time the Photograph is Taken

The opportunity to retake a bad photograph (photo) is often short-lived. The photographer will likely not be in the same place again under the same conditions, in order to be able to retake a photo to improve the quality. If the photo is sufficiently bad, then the photographer may discard and retake the photo. However, with the advent of photo sharing over social media, the quality of the photo itself is often established by discussion and liking of a group of contacts with whom the user shared the photo .

The novel system uses machine-learning techniques to establish a base set of criteria for the quality of photos that the user may like. These criteria are based on photos that:

Others posted on social media and the user liked
The user posted on social media and others liked

The system applies this set of criteria, in the form of a statistical model, to score the quality of the photo at the time the user takes it. If the photo does not meet a predefined threshold in quality, then the system prompts the user to keep, retake, or discard the photo.

This provides an advanced photo quality feature that is useful for smart mobile devices with both a camera and access to social media applications. This feature goes beyond the existing art because it enables the recognition of a photo as high quality both at the time of capture and based on the likelihood that it scores highly amongst a peer group in social media .

The novel system uses existing machine learning techniques to build a statistical model of the user's photos in social media. Existing statistical model techniques work well because said techniques excel at establishing a scoring mechanism over a broad set of features that are trained on a set of positive and negative examples . Many features of the photos can be measured for inclusion in the statistical model generation, such as (but not limited to):

Color depth, Complexity Number of people identified Other recognized objects in the photos (ex: sports cars)

Ratio of friends/non-friends in social media Inclusion of a celebrity...