Browse Prior Art Database

Method and apparatus for sensitive image assessment Disclosure Number: IPCOM000243727D
Publication Date: 2015-Oct-15
Document File: 2 page(s) / 58K

Publishing Venue

The Prior Art Database


We propose a sensitive images assessment method. It automatically evaluates the sensitivity of a given image which is typically taken by people by their cameras. The idea is using both local visual feature and global layout of the local features to reason the sensitivity of a given image. This method can be applied in many cases such as helping the user to decide if the image in his/her cell phone is very sensitive and is not suggested to upload to a third-party cloud environment.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 80% of the total text.

Page 01 of 2

Method and apparatus for sensitive image assessment

Images are becoming ubiquitous due to massive mobile devices and their associated digital cameras. Moreover, the battery and data storage capability are both growing thus making the images become more and more for individual users. It also brings about risk because some images may contain very sensitive personal information, such as his ID number, his bank account number, and even some password. On the other hand, some pornographic images cause embarrassment and do harm to the society. For the above problem, we aim to present a method for automatically evaluating the sensitivity of the images, based on their visual content.

Our approach consists of the following components:
1) Components detector, which performs face, text, serial detection and recognition to test the presence of each component
2) Components aggregation and association scorer, which

2.1 Train or define an association model (template)

2.2 Apply the model to evaluate the overall score for the detected components

2.3 Specified personal and customized sensitive information can be further highlighted given the personal sensitive information database.

3) Automatic generated association relation template (by clustering)

3.1 Input images labelled as sensitive photos
3.2 Detect the components using the component detector
3.3 Cluster the typical layout and/or co-existence of multiple components as association templates
4) Alert and active management module vi...