Implementation of video tagging identifying characters in video
2 likes267 views
The document presents a method for automatic character identification in videos through video tagging, utilizing face detection and recognition techniques. By employing the Emgu CV image processing library, the approach captures and labels faces in motion, storing their identities for future reference. The system is designed to enhance content accessibility on platforms like social media and improve biometric security measures without the need for individuals to stop for identification.
![IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
__________________________________________________________________________________________
Volume: 03 Issue: 05 | May-2014, Available @ http://www.ijret.org 183
3. WORK DONE
The idea in this paper uses the above mentioned Emgu, an image
processing library to achieve the video tagging concept. The
process consists of mainly three following steps:
When the video is played or streamed the faces appearing on the
screen will be detected. Faces will have different orientation, size
and color. Only after the face of a person is detected, the
information of its identity will be entered that can be Name, ID,
Gender etc. of the person. This information will be stored in the
database as a tag which will be called in the future reference. So
when a video is played and the same person whose information is
already stored appears in that video, the system will recognize
him [8] and displays his related information.
4. IMPLEMENTATION OF VIDEO TAGGING
4.1 Detecting the Face
It uses simple rectangular features also called Haarfeature[2],
where each feature is a single valueobtained bysubtracting sum
of pixels under white rectangle from sumof pixels under black
rectangle. If the difference is above threshold, corresponding
feature like eyes, nose etc. are present. Then a special
representation of the sample called the integral image is used
which makes feature extraction faster. A series of AdaBoost[2]
classifiers works as a filter chain. Image sub regions that make it
through the entire cascade are classified as face. All others are
classified as non-face. The classifiers are used efficiently by
assigning heavily weighted filters first on the window, if a
window fails the first stage, discard it and we don’t consider
remaining features on it. If it passes we apply the remaining
featuresand continue the process. Finally once the face is
detected using above filters, we draw a focus around the region
to highlight the face.
4.2 Recognizing the face:
Once the faces are detected they are saved into database as
image templates. All the trained faces are histogram normalized
to convert them into 0-255 gray scale [3] and correspondingly
object recognizer is created for the given faces. Calculation of
Eigen image and correspondingly Eigen values are done after
that.And when the new face is detected, calculation of Eigen
distance value [4] of the projected face with the trained faces
isperformed.If the value of Eigen distance is greater than
threshold for a particular image, and we display the label of the
correspondingtrained face to the projected face.
Fig 2 Video tagging illustrations](https://image.slidesharecdn.com/implementationofvideotaggingidentifyingcharactersinvideo-140819062233-phpapp02/85/Implementation-of-video-tagging-identifying-characters-in-video-2-320.jpg)
![IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
__________________________________________________________________________________________
Volume: 03 Issue: 05 | May-2014, Available @ http://www.ijret.org 184
5. CONCLUSIONS
Video tagging is done on a video with the concept explained in
the paper. Figure 2 shows the process steps taking place, a face
is recognized successfully after matching with the image
template of the person stored for the first time. Now if the same
person appears in that video or any other video, the face will be
automatically recognized by the system with a tag to tell who the
person is. Implemented concept of video tagging in this paper
can have various utilities.
FUTURE WORK
The basic concept will remain the same which is to identify a
character present in the video [5] or live streaming. The
following work can be done in future based on the work done-
Tagging a Video on the Facebook
Photo tagging is a major reason that makes photos special on
Facebook. So video tagging can be introduced too. You can tag a
video to show who’s in the video or tell who you are with. The
tag will be the user name of the person. When a friend is tagged
in the video, he will be notified through an alert. The user needs
to be tagged just once in a video and every time that person
appears in any video uploaded later; he will be recognized
automatically like auto photo tagging.
Biometric Security Checks for Secure Areas
In normal face recognition biometric test a person stands in front
of the detector and matches the traits with the once stored in the
system. In that a person should be still to match the features. So
using this idea there is no need to stop and unlock. The person
can be recognized while in motion and the camera will detect
and recognize if the person appearing on stream is valid in
database or not and permission will be granted if match is found.
REFERENCES
[1]. Jitao Sand and ChangshengXu, “Robust Face-Name Graph
Matching for Movie Character Identification”,
IEEETransactionson multimedia Vol. 14, no3, June 2012
[2]. P. Viola and M Jones, “Rapid Object Detection using
boosted cascade of simple features”. Computer Vision and
Pattern Recognition, IEEE CVPR.2001.990517 Vol1.
[3]. Dalal N and Triggs B, “Histograms of oriented gradients for
human detection”.Computer Vision and Pattern Recognition,
IEEE CVPR.2005.177 Vol1.
[4]. A. Sanfeliu and K. Fu, “A distance measure between
attributed relationalgraphs for pattern recognition,” IEEE Trans.
Syst. Man Cybern.vol.13, no. 3, 1983.
[5]. J. Stallkamp, H. K. Ekenel, and R. Stiefelhagen, “Video-
based face recognition on real-world data,” in Proc. Int. Conf.
Comput. Vis. 2007, pp. 1-8
[6]. channel9.msdn.com/Face-Detection-with-Emgu-CV-in-C-
and-WPF.
[7]. M. DharmatejaPurna, N. Praveen “A Novel Method for
Movie Character Identification and its Facial Expression
Recognition”, IJMER Vol.3, Issue.3 2013
[8]. W. Zhao, R. Chelappa, P. J. Phillips, and A. Rosenfeld,
“Face recognition: A literature survey,” ACM Compu. Survey.,
vol. 35, no. 4, pp. 399–458, 2003](https://image.slidesharecdn.com/implementationofvideotaggingidentifyingcharactersinvideo-140819062233-phpapp02/85/Implementation-of-video-tagging-identifying-characters-in-video-3-320.jpg)
Implementation of video tagging identifying characters in video
- 1. IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308 __________________________________________________________________________________________ Volume: 03 Issue: 05 | May-2014, Available @ http://www.ijret.org 182 IMPLEMENTATION OF VIDEO TAGGING: IDENTIFYING CHARACTERS IN VIDEO Akshay Bhardwaj1 , Ramani S2 , Mayank Goyal3 1 Student, School of Computing Science and Engineering, Vellore Institute of Technology, Vellore, India 2 Assistant Professor, School of Computing Science and Engineering, Vellore Institute of Technology, Vellore, India 3 Student, School of Computing Science and Engineering, Vellore Institute of Technology, Vellore, India Abstract Automatic face detection in movies has been an interesting area of research. Due to the various appearances of the characters in the moving film it is not easy to identify the true faces. There has been a significant growth in the field of image processing to detect still human faces with high accuracy and results. This idea can be further used to implement video tagging. We see biometric security check points that compare still images to verify a person in the database and the same can be done if a person is in the motion. This paper proposes a method to identify characters appearing in the moving film. Keywords—Video Tagging, Image processing, Eigen objects, Security, Template matching ----------------------------------------------------------------------***---------------------------------------------------------------------- 1. INTRODUCTION The growing video and TV content on the internet comes with huge amount of digital video data just like still images. This resulted in studying and developing efficient techniques for understanding the video content. The concept of automatic video analysis is one of the technologies which is studied and implemented here. Video tagging is a concept similar to image tagging with a difference being that the faces are in the motion in the former. The key concept is to store the image from a running video and storing it after training under constraint like size of image, gray scale and bytes. Photo tagging has been implemented in various social media such as Facebook and Instagram where you can tag a person once and the next time you upload a photo of the same character his name will appear automatically. The objective is to identify the human faces present in the running film and label them with their corresponding identity for example the name of character. Figure1 explains the underlying idea by pointing the name of faces detected. Fig.1 2. RELATED WORK Template matching is a common technique where an object is captured as an image called template. The match will fail if the object appears scaled, rotated, or skewed on the image. Instead, this can be done using Emgu CV image processing library.It is written entirely in C#. The benefit is that it can be compiled in mono and therefore is able to run on any platform Mono supports, including Linux, Mac OS X, win. It comprises of 2 layers, Layer 1 contains functions, enumeration, structures and Layer 2 has classes. The figure below is the architecture overview of the Emgu CV library.
- 2. IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308 __________________________________________________________________________________________ Volume: 03 Issue: 05 | May-2014, Available @ http://www.ijret.org 183 3. WORK DONE The idea in this paper uses the above mentioned Emgu, an image processing library to achieve the video tagging concept. The process consists of mainly three following steps: When the video is played or streamed the faces appearing on the screen will be detected. Faces will have different orientation, size and color. Only after the face of a person is detected, the information of its identity will be entered that can be Name, ID, Gender etc. of the person. This information will be stored in the database as a tag which will be called in the future reference. So when a video is played and the same person whose information is already stored appears in that video, the system will recognize him [8] and displays his related information. 4. IMPLEMENTATION OF VIDEO TAGGING 4.1 Detecting the Face It uses simple rectangular features also called Haarfeature[2], where each feature is a single valueobtained bysubtracting sum of pixels under white rectangle from sumof pixels under black rectangle. If the difference is above threshold, corresponding feature like eyes, nose etc. are present. Then a special representation of the sample called the integral image is used which makes feature extraction faster. A series of AdaBoost[2] classifiers works as a filter chain. Image sub regions that make it through the entire cascade are classified as face. All others are classified as non-face. The classifiers are used efficiently by assigning heavily weighted filters first on the window, if a window fails the first stage, discard it and we don’t consider remaining features on it. If it passes we apply the remaining featuresand continue the process. Finally once the face is detected using above filters, we draw a focus around the region to highlight the face. 4.2 Recognizing the face: Once the faces are detected they are saved into database as image templates. All the trained faces are histogram normalized to convert them into 0-255 gray scale [3] and correspondingly object recognizer is created for the given faces. Calculation of Eigen image and correspondingly Eigen values are done after that.And when the new face is detected, calculation of Eigen distance value [4] of the projected face with the trained faces isperformed.If the value of Eigen distance is greater than threshold for a particular image, and we display the label of the correspondingtrained face to the projected face. Fig 2 Video tagging illustrations
- 3. IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308 __________________________________________________________________________________________ Volume: 03 Issue: 05 | May-2014, Available @ http://www.ijret.org 184 5. CONCLUSIONS Video tagging is done on a video with the concept explained in the paper. Figure 2 shows the process steps taking place, a face is recognized successfully after matching with the image template of the person stored for the first time. Now if the same person appears in that video or any other video, the face will be automatically recognized by the system with a tag to tell who the person is. Implemented concept of video tagging in this paper can have various utilities. FUTURE WORK The basic concept will remain the same which is to identify a character present in the video [5] or live streaming. The following work can be done in future based on the work done- Tagging a Video on the Facebook Photo tagging is a major reason that makes photos special on Facebook. So video tagging can be introduced too. You can tag a video to show who’s in the video or tell who you are with. The tag will be the user name of the person. When a friend is tagged in the video, he will be notified through an alert. The user needs to be tagged just once in a video and every time that person appears in any video uploaded later; he will be recognized automatically like auto photo tagging. Biometric Security Checks for Secure Areas In normal face recognition biometric test a person stands in front of the detector and matches the traits with the once stored in the system. In that a person should be still to match the features. So using this idea there is no need to stop and unlock. The person can be recognized while in motion and the camera will detect and recognize if the person appearing on stream is valid in database or not and permission will be granted if match is found. REFERENCES [1]. Jitao Sand and ChangshengXu, “Robust Face-Name Graph Matching for Movie Character Identification”, IEEETransactionson multimedia Vol. 14, no3, June 2012 [2]. P. Viola and M Jones, “Rapid Object Detection using boosted cascade of simple features”. Computer Vision and Pattern Recognition, IEEE CVPR.2001.990517 Vol1. [3]. Dalal N and Triggs B, “Histograms of oriented gradients for human detection”.Computer Vision and Pattern Recognition, IEEE CVPR.2005.177 Vol1. [4]. A. Sanfeliu and K. Fu, “A distance measure between attributed relationalgraphs for pattern recognition,” IEEE Trans. Syst. Man Cybern.vol.13, no. 3, 1983. [5]. J. Stallkamp, H. K. Ekenel, and R. Stiefelhagen, “Video- based face recognition on real-world data,” in Proc. Int. Conf. Comput. Vis. 2007, pp. 1-8 [6]. channel9.msdn.com/Face-Detection-with-Emgu-CV-in-C- and-WPF. [7]. M. DharmatejaPurna, N. Praveen “A Novel Method for Movie Character Identification and its Facial Expression Recognition”, IJMER Vol.3, Issue.3 2013 [8]. W. Zhao, R. Chelappa, P. J. Phillips, and A. Rosenfeld, “Face recognition: A literature survey,” ACM Compu. Survey., vol. 35, no. 4, pp. 399–458, 2003