Face Tracking in the Compressed Domain

Fonseca, Pedro; Nesvadba, J.

doi:10.1155/asp/2006/59451

Cited by 7 publications

(8 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, work [2]'s performance is not tested in common and publicly available content. Hence, we first compared our algorithm with forward MVs-based algorithm (work [6] used forward and backward MVs-based algorithm that is not strictly forward MVs-based algorithm). Table III illustrates that forward MVs-based algorithm produces recall value of 87% and precision value of 85%.…”

Section: Experiments Resultsmentioning

confidence: 99%

“…The proposed algorithm improves about 7% of recall and about 8% of precision. Previous work [6] performs in MPEG1/MPEG2 compressed domain. Our results have been compared with them.…”

Section: Experiments Resultsmentioning

confidence: 99%

“…It implies that two observation regions rely on frames more than one. Different from work [6], we use all reference frames for tracking faces. However, emphasize that size of buffers dose not increase, because H.264/AVC provides buffer storages for reference frames already.…”

Section: B Face Tracking Algorithmmentioning

confidence: 99%

“…These learning-based approaches not only involve much trouble in that had to collect manually a huge image database, but yield the lower accuracy in detecting the smaller faces such as QCIF video sequences. P. Fonseca and J. Nesvadba [5,6] proposed a face detection system, in which frontal and non-frontal faces are detected by using information derived from color, DC coefficients and AC coefficients in MPEG1/MPEG2 video sequences. However, DC image obtained based DC coefficients is too small to apply to detecting small faces in QCIF video sequences.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Face track in H.264 compressed domain using the Face Observation-MRF model

Nam

Ruan

2010

IEEE 10th INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING PROCEEDINGS

View full text Add to dashboard Cite

In this paper, we propose a face track algorithm to determine the presence and location of faces of sizes more than 15×15 pixels in all frames of QCIF video sequences encoded by H.264/AVC. The proposed system composes of detection and tracking algorithms that all perform in H.264/AVC JM15.0 Encoder. The location of multiple frontal and non-frontal faces in compressed video sequences could be detected using information derived from YCbCr and AC coefficients that are taken from Iframe encoding mode. The face detection algorithm is combined with face track algorithm that performs face tracking in Pframes of compressed video sequences by using the Face Observation-based Markovian Random Field model. The proposed algorithm is not only affected less from noise MVs and it's motion changes, but also offers low computational cost solution to problem of tracking faces. Experiments show that our algorithm provides the remarkable performance and can extract the location of faces on H.264/AVC compressed domain efficiently and robustly.

show abstract

Section: Experiments Resultsmentioning

confidence: 99%

“…The proposed algorithm improves about 7% of recall and about 8% of precision. Previous work [6] performs in MPEG1/MPEG2 compressed domain. Our results have been compared with them.…”

Section: Experiments Resultsmentioning

confidence: 99%

Section: B Face Tracking Algorithmmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Face track in H.264 compressed domain using the Face Observation-MRF model

Nam

Ruan

2010

IEEE 10th INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING PROCEEDINGS

View full text Add to dashboard Cite

show abstract

“…However, the majority of these techniques deal with only one class of skin colour [7,8,6]. Faces can be tracked using a Kalman filtering technique to model face trajectories [20,4,13] with application for example in the compressed video domain. For instance, in [20] people need to be facing the camera in a closed up view for the system to be able to extract eyes and mouth locations as tracking inputs.…”

Section: State Of the Artmentioning

confidence: 99%

Combining face detection and people tracking in video sequences

Corvee¹,

Brémond²

2009

3rd International Conference on Imaging for Crime Detection and Prevention (ICDP 2009)

View full text Add to dashboard Cite

Face detection algorithms are widely used in computer vision as they provide fast and reliable results depending on the application domain. A multi view approach is here presented to detect frontal and profile pose of people face using Histogram of Oriented Gradients, i.e. HOG, features. A K-mean clustering technique is used in a cascade of HOG feature classifiers to detect faces. The evaluation of the algorithm shows similar performance in terms of detection rate as state of the art algorithms. Moreover, unlike state of the art algorithms, our system can be quickly trained before detection is possible. Performance is considerably increased in terms of lower computational cost and lower false detection rate when combined with motion constraint given by moving objects in video sequences. The detected HOG features are integrated within a tracking framework and allow reliable face tracking results in several tested surveillance video sequences.

show abstract

A DCT-Based Approach for Hiding Patients' Identification Information

Popa

Pop

Nemes

et al. 2008

2008 IEEE International Conference on Automation, Quality and Testing, Robotics

View full text Add to dashboard Cite

An algorithm for hiding patient's data from JPEG ultrasound images, applied directly on DCT (Discrete Cosine Transform) coefficients is presented. The algorithm can detect textual information using the amount of energy, computed using only AC coefficients, without converting medical images/video back to the spatial (uncompressed) domain. In order to offer a high quality care for patients, textual information containing the patient data is hidden or eliminated. Further, the processed medical images can be accessed and analyzed, for a second-opinion, by physicians, or in medical research, keeping the patients privacy provided of the ultrasound medical images. The algorithm has been implemented and verified with good performances.

show abstract

Face Tracking in the Compressed Domain

Cited by 7 publications

References 13 publications

Face track in H.264 compressed domain using the Face Observation-MRF model

Face track in H.264 compressed domain using the Face Observation-MRF model

Combining face detection and people tracking in video sequences

A DCT-Based Approach for Hiding Patients' Identification Information

Contact Info

Product

Resources

About