Logotypes detection and other characteristic invariants
From LRDE
- Authors
- Anthony Seure
- Type
- techreport
- Year
- 2013
- Number
- 1308
- Projects
- Olena
- Keywords
- C++, Olena, invariants, descriptor, SIFT
Abstract
The detection of logotypes and invariants inside an image aims to find within an image (or a sequence of images) a new or already known graphical element which pertains to a brand, a company, somebody, etc. Such elements could be found in many real-life pictures but also in advertising pictures. The detection challenge is to compare elements with those which have been already seen but also to set up a machine learning approach in order to determine if there are new logotypes. The integration of such a tool inside the Olena image processing platform and possibly inside the Terra Rush project could lead to a better content indexing and to invalidate specific areas in other processing toolchains. In this report, we will mainly explain a generic method to locate invariants keypoints of an image : the SIFT descriptor.