Improving Face Detection Performance of Compressed MPEG Videos by Using Frame-Independent Scene Change Detection Method

Abstract

With the spread of computer vision applications, the performance of such applications also became prominent, especially when real-time and near real-time use cases are considered. If not all, many object detection algorithms follow a frame-based search approach, where all frames of the MPEG stream are analyzed sequentially. This drastically increases the computation time and the hardware requirements for such systems. This paper proposes employing a new scene-change detection method to improve object and face detection performance by eliminating the need to analyze every video frame. The method provides a frameindependent approach and does not require decoding and reencoding of MPEG video. The paper also reports the performance test outcomes to exhibit the proposed approach’s value. The findings show that a scene-change detection method enhances efficiency and decreases computational demands. Focusing on frames that show scene changes shows notable advancements in object detection performance.

Keywords

• Face detection
• Object detection
• Frame-by-frame analysis

References

1. [1] E. Hjelmas and B. Low, “Face detection: A survey,” ˚ Computer Vision and Image Understanding, vol. 83, no. 3, September 2001.
2. [2] L. Jiao, R. Zhang, F. Liu, S. Yang, B. Hou, L. Li, and X. Tang, “New generation deep learning for video object detection: A survey,” IEEE Transactions on Neural Networks and Learning Systems, vol. 33, no. 8, August 2022.
3. [3] L. Baraldi, C. Grana, and R. Cucchiara, “Measuring scene detection performance,” in Pattern Recognition and Image Analysis: Proceedings of 7th Iberian Conference, IbPRIA 2015, Santiago de Compostela, Spain, June 2015.
4. [4] I. Sethi and N. Patel, “A statistical approach to scene change detection,” in Proceedings of SPIE - The International Society for Optical Engineering, June 1998.
5. [5] C.-L. Huang and B.-Y. Liao, “A robust scene-change detection method for video segmentation,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 11, no. 12, December 2001.
6. [6] B. Shahraray, “Scene change detection and content-based sampling of video sequences,” in Proceedings Volume 2419, Digital Video Compression: Algorithms and Technologies, April 1995.
7. [7] R. Das¸, B. Polat, and G. Tuna, “Recognizing and tracking objects in images and videos with deep learning,” Fırat University, Muhendislik ¨ Bilimleri Dergisi, vol. 31, no. 2, pp. 571–581, 2019.
8. [8] Bitmovin, “Video developer report 2018,” September 2019.

9. [9] A. Huszak and S. Imre, “Analysing gop structure and packet loss effects on error propagation in mpeg-4 video streams,” in Proceedings of 4th International Symposium on Communications Control and Signal Processing (ISCCSP), March 2010.
10. [10] B. Zatt, M. Porto, J. Scharcanski, and S. Bampi, “Gop structure adaptive to the video content for efficient h.264/avc encoding,” in Proceedings of IEEE International Conference on Image Processing, September 2010.
11. [11] Z. Wang, A. Bovik, H. Sheikh, and E. Simoncelli, “Image quality assessment: From error visibility to structural similarity,” IEEE Transactions on Image Processing, April 2004.
12. [12] Z. Wang and Q. Li, “Information content weighting for perceptual image quality assessment,” IEEE Transactions on Image Processing, May 2011.