Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

A Method and Apparatus of Fast Trademark Candidate Localization in Stream Media

IP.com Disclosure Number: IPCOM000196566D
Publication Date: 2010-Jun-07
Document File: 2 page(s) / 103K

Publishing Venue

The IP.com Prior Art Database

Abstract

In modern society, advertisement becomes more and more important for a company or a specific product. A lot of money has been spent on ads every year. For example, Apple has already spent about 500 million dollars for ads this year; and Microsoft has spent 1.4 billion dollars. More specifically, a big proportion of this money has been spent on stream media ads, such as TV ads and internet stream media ads. Companies usually want to evaluate the effect of their ads, and consequently adjust their strategy of advertisement. To fulfill this task, they need to recognize their ads from media. Before recognizing they should find some features that can stand for the ads in stream media. Thus there raise a problem that how to detect the ads in stream media. Existing video retrieval is a straight-forward solution, but its performance is still far from practical. So the key problems are clear: How to fast, automatically and precisely position the trademark in stream media? We here address the problem as trademark localization in stream media. By investigating large numbers of ads in stream media, we found that the trademark for a company or a specific product always show up in advertisement, and it always stay still for quite a while so that audiences can see it clearly. Thus we invent a method to efficiently detect the trademarks in stream media.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 100% of the total text.

Page 1 of 2

A Method and Apparatus of Fast Trademark Candidate Localization in Stream Media

Architecture (Fig 1):

Followings are the key components and their functions in this method:

Fig.1. The whole process of the invention.

Gappy Frame Differential (Fig.2)

A component to create the differential between adjacent frames and gappy frames.

Fig.2. Gappy frame differential.

Implementation details:
Given a stream media V , the gappy differential '

V can be computed as follows:

n

F

(

k

)

=

(

f

(

k

)

f

(

k

i

))

i

=

1

Where n is the length of gappy window, )

f is a frame of V , )

(*

F is the corresponding frame of '

V . k is the

(*

subscript of frames.

Still Region Detection (Fig.3)

A component to detect the still regions in stream media V .

Implementation details
First, we detect the zeros in gappy frame differential. If most pixels in a region are zeros, we expect this region which stays unchanged in a time window is a still region. Second, a typical frame in the time window is chosen, and this still region can be represented as corresponding part of this typical frame.

1

[This page contains 2 pictures or other non-text objects]

Page 2 of 2

Fig.3 Still region detection.

Still Region Entropy


A component to calculate the entropy of the still region thus filter out some false positives. Implementation details
This can be done by calculating the variance of the still region.

2

[This page contains 1 picture or other non-text object]