DOI QR코드

DOI QR Code

Learning-based Super-resolution for Text Images

글자 영상을 위한 학습기반 초고해상도 기법

  • Received : 2014.12.12
  • Accepted : 2015.03.26
  • Published : 2015.04.25

Abstract

The proposed algorithm consists of two stages: the learning and synthesis stages. At the learning stage, we first collect various high-resolution (HR)-low-resolution (LR) text image pairs, and quantize the LR images, and extract HR-LR block pairs. Based on quantized LR blocks, the LR-HR block pairs are clustered into a pre-determined number of classes. For each class, an optimal 2D-FIR filter is computed, and it is stored into a dictionary with the corresponding LR block for indexing. At the synthesis stage, each quantized LR block in an input LR image is compared with every LR block in the dictionary, and the FIR filter of the best-matched LR block is selected. Finally, a HR block is synthesized with the chosen filter, and a final HR image is produced. Also, in order to cope with noisy environment, we generate multiple dictionaries according to noise level at the learning stage. So, the dictionary corresponding to the noise level of the input image is chosen, and a final HR image is produced using the selected dictionary. Experimental results show that the proposed algorithm outperforms the previous works for noisy images as well as noise-free images.

본 논문은 글자 영상을 효과적으로 확대 (up-scaling)하기 위한 학습 기반 초고해상도 (super-resolution; SR) 기법을 제안한다. 제안 기법은 크게 학습 단계와 합성 단계로 나뉜다. 학습 단계에서 다양한 HR (high-resolution) /LR (low-resolution) 글자 영상 쌍들을 수집한다. LR영상들은 양자화를 하고, 충분히 많은 수의 HR-LR 블록쌍들을 추출한다. 양자화된 LR블록을 기준으로 블록 쌍들을 소정의 개수의 클래스들로 구분한다. 클래스 별로 최적의 2D-FIR 필터 계수를 계산하고, 양자화한 후색인용 LR 블록과 함께 사전에 저장한다. 합성 단계에서 입력 LR 영상 내 각 블록을 양자화한 후 사전 내 양자화된 LR블록들과 정합하여 가장 근사한 블록에 대응하는 FIR 필터계수를 선정한다. 마지막으로 선택된 FIR필터로 HR 블록을 합성하여 최종적인 HR영상을 생성한다. 또한, 우리는 잡음이 있는 글자 영상에 대응하기 위해 학습과정에서 잡음 세기에 따른 복수개의 사전들을 제작한다. 입력 LR 영상의 잡음 레벨에 맞는 사전을 선택하여 HR영상을 합성한다. 실험 결과는 제안 기법이 종래 기법보다 잡음이 없는 환경에서는 물론 잡음이 있는 환경에서 우수한 주관적/객관적 화질을 가짐을 보인다.

Keywords

References

  1. W. Freeman, T. R. Jones, and E C. Pasztor. "Example-based super-resolution." IEEE Computer Graphics and Applications, vol. 22, no. 2, pp. 56-65, Oct. 2002 https://doi.org/10.1109/38.988747
  2. J. Park, Y. Kwon, and J. H. Kim, "An example-based prior model for text image super-resolution," Proc. IEEE ICDAR, pp. 374-378, 2005.
  3. K. I. Kim, and Y. H. Kwon. "Example-based learning for single-image super-resolution," Pattern Recognition, pp. 456-465, 2008.
  4. R. P. Duda, P. E. Hart, D. G. Stork, "Pattern Classification 2/e", WILEY-INTERSCIENCE, pp. 526-530.
  5. H. Anton, R. C. Busby, "Contemporary linear algebra," WILEY, pp. 51-53.
  6. H. Luong and W. Philips, "Robust reconstruction of low-resolution document images by exploiting repetitive character behaviour," International Journal of Document Analysis and Recognition (IJDAR), vol. 11, no. 1, pp. 39-51, 2008. https://doi.org/10.1007/s10032-008-0068-2
  7. C. E. Duchon, "Lanczos filtering in one and two dimensions," Journal of Applied Meteorology, vol. 18, pp. 1016-1022, August 1979. https://doi.org/10.1175/1520-0450(1979)018<1016:LFIOAT>2.0.CO;2
  8. D. Kostadin et al., "BM3D image denoising with shape-adaptive principal component analysis." Signal Processing with Adaptive Sparse Structured Representations, 2009.
  9. S. Yang, M. Wang, Y. Chen, Y. Sun, "Single-image super-resolution reconstruction via learned geometric dictionaries and clustered sparse coding," IEEE Trans. Image Process., vol. 21, no. 9, 2012.
  10. T. M. Chan et al. "Neighbor embedding based super-resolution algorithm through edge detection and feature selection," Pattern Recognition Letters, pp. 494-502, 2009.
  11. H. Luong, W. Philips, "Non-local text image reconstruction," Proc. ICDAR, vol. 1, pp. 546-550, 2007.
  12. R. Walha, F. Drira, F. Lebourgeois, A. M. Alimi, "Super-resolution of single text image by sparse representation," Proc. ICVGIP, pp. 22-29, 2012.