Application of fractal analysis in image motion estimation

ABSTRACT

The motion estimation, as a key technique, affects the effect of image processing. However, many matching criteria based on the similarity of image grey value probably lead to the inaccurate motion estimation. In order to improve the performance of motion estimation, a novel image matching criterion is proposed on the basis of the fractal theory. The proposed criterion not only considers the grey-level similarity of the same object, but also takes into account the invariance of the fractal dimension of the same object. Therefore, the optimal and accurate matching between anchor object and reference objects can be better ensured. The experimental results show that the estimated motion vector of the proposed method is more approximate to the true motion vector than that of the traditional methods. What is more, the peak signal-to-noise ratio value of the image is improved by the proposed criterion. Meantime, the computational complexity is increased slightly.

KEYWORDS:

• Fractal analysis
• motion estimation
• objective function
• matching criterion
• matching block

Notes

The American football image can be accessed through the following URL: http://www.cipr.rpi.edu/resource/sequences/index.html, courtesy of CIPR test image sequences.

The Foreman and Bus images can be accessed through the following URL: http://trace.eas.asu.edu/, courtesy of Arizona State University Video Trace Library.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Zhiyong Shi is an assistant professor at Chongqing Communication Institute, Chongqing, China. He is currently pursuing his Ph.D. degree at the College of Communication Engineering, Chongqing University, Chongqing, China. He received his BS degree in Communication Engineering from Chongqing Communication Institute, Chongqing, China, in 2000, and his MS degree in Communication and Information System from South China University of Technology, Guangzhou, China, in 2006. He is the author of more than 20 journal papers and has written one book chapter. His current research interests include data image processing and mobile network technique.

Fengchun Tian is now a professor, Ph.D. supervisor of Chongqing University, P.R. China. He received his BS, MS and Ph.D. degrees from Chongqing University, Chongqing, China, in 1984, 1989 and 1996, respectively. He is the author of more than 90 journal papers and has written four book chapters. His research interests include image processing, intelligent algorithms, wavelet theory and its application, fractal and chaos, optical information processing, and electronic nose for odor detection.

Yande Wang is a researcher at Chongqing Communication Institute, Chongqing, China. His current research interests include data image processing and mobile network technique.

Jian Ran is currently pursuing his MS degree at the College of Communication Engineering, Chongqing University, Chongqing, China. His research interests include frames rate up-conversion and image processing.

Additional information

Funding

This work was jointly supported by the National Natural Science Foundation of China [number 61201347], Chongqing foundation and advanced research project [number cstc2016jcyjA0103]. The support is gratefully acknowledged.