DOI QR코드

DOI QR Code

Super Resolution Technique Through Improved Neighbor Embedding

개선된 네이버 임베딩에 의한 초해상도 기법

  • Eum, Kyoung-Bae (Kunsan University, Dept. of Computer and Information Eng.)
  • Received : 2014.12.08
  • Accepted : 2014.12.31
  • Published : 2014.12.31

Abstract

For single image super resolution (SR), interpolation based and example based algorithms are extensively used. The interpolation algorithms have the strength of theoretical simplicity. However, those algorithms are tending to produce high resolution images with jagged edges, because they are not able to use more priori information. Example based algorithms have been studied in the past few years. For example based SR, the nearest neighbor based algorithms are extensively considered. Among them, neighbor embedding (NE) has been inspired by manifold learning method, particularly locally linear embedding. However, the sizes of local training sets are always too small. So, NE algorithm is weak in the performance of the visuality and quantitative measure by the poor generalization of nearest neighbor estimation. An improved NE algorithm with Support Vector Regression (SVR) was proposed to solve this problem. Given a low resolution image, the pixel values in its high resolution version are estimated by the improved NE. Comparing with bicubic and NE, the improvements of 1.25 dB and 2.33 dB are achieved in PSNR. Experimental results show that proposed method is quantitatively and visually more effective than prior works using bicubic interpolation and NE.

단일 영상 초해상도 기법에는 보간 기반 방법과 표본 기반 방법 등이 있다. 보간 기반 방법들은 간결성에 강점을 가지고 있으나, 이들 방법들은 선지식을 이용할 수 없기 때문에 톱니 모양의 윤곽선을 가진 고해상도 영상을 생성하는 경향이 있다. 표본 기반 초해상도 기법에서는 최근방 기반 알고리즘들이 널리 이용되어 지고 있다. 그들 중, 네이버 임베딩은 지역적 선형 임베딩이라는 매니폴드 학습 방법의 개념과 같다. 그러나, 네이버 임베딩은 국부 학습 데이터 집합의 크기가 너무 작은데에 따른 빈약한 일반화 능력으로 인하여, 시각적으로나 정량적인 척도에 의해 취약한 성능을 보인다. 본 논문에서는 이와 같은 문제점을 해결하기 위해 개선된 네이버 임베딩 알고리즘을 제안하였다. 저해상도 입력 영상이 주어지면 고해상도 버전의 화소 값들은 개선된 네이버 임베딩 알고리즘에 의해 구해진다. 실험 결과 제안된 방법이 바이큐빅 보간법이나 네이버 임베딩에 비해 정량적인 척도 및 시각적으로도 우수한 결과를 보였다.

Keywords

References

  1. R. S. Wagner, D. E. Waagen, and M. L. Cassabaum, "Image super resolution for improved automatic target recognition," in Proc. SPIE, vol. 5426, 2004.
  2. D. Li and Steven Simske, "Example based single fra me image super resolution by support vector regression," Patten Recognition Research, vol. 5, no. 1, 2010.
  3. J. van Ouwerkerk, "Image super resolution survey," Image and Vision Computing, vol. 24, no. 10, 2006.
  4. W. Freeman, T. Jones, and E. Pasztor, "Example based super resolution," IEEE Computer Graphics and Applications, vol. 22, no. 2, 2002.
  5. Yi Tang and Xuelong Li, "Local semi-supervised regression for single image super resolution," IEEE Int. Workshop on MMSP, 2011.
  6. H. Chang, D. Yeung, and Y. Xiong, "Super resolution through neighbor embedding," in Proc. IEEE CVPR, vol. 1, 2004.
  7. M. Bevilacqua, A. Roumy, and C. Guillemot, "Super-resolution using Neighbor Embedding of Back-projection residuals," in 18th International Conference on Digital Signal Processing, 2013.
  8. Jili G Vadukkoot and Tinu Winson K, " A new image resolution technique using neighbor embedding," Int. Journal of Advances in Computer Science and Technology, vol. 3, no. 2, 2014.
  9. N. Cristianini and J. Shawe-Taylor, "An introduction to Support Vector Machines," New York:Cambridge Univ. Press, 2000.
  10. V. Vapnik, "The nature of statistical learning theory," Springer Verlag, New York, 1995.
  11. C. Chang and C. Lin, LIBSVM : a library for support vector machines, 2001.