Super-resolution reconstruction with non sub-sampled contourlet transform and image self-similarity
| Autor: | Chenghui Liu, Binwen Fan, Tiantian Zhang, Bo Wu |
|---|---|
| Rok vydání: | 2017 |
| Předmět: |
Computer science
business.industry Feature extraction ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Pattern recognition Iterative reconstruction Sparse approximation Superresolution Contourlet Image (mathematics) Feature (computer vision) Computer Science::Computer Vision and Pattern Recognition Artificial intelligence business Image resolution Interpolation |
| Zdroj: | 2017 IEEE 17th International Conference on Communication Technology (ICCT). |
| DOI: | 10.1109/icct.2017.8359900 |
| Popis: | Image super-resolution based on sparse representation is proved to be a promising single-image super-resolution technology that trains the external dictionaries by extracting the feature representations of images to restore the lost high frequency information of the input image. To improve the quality of the reconstructed image, this paper proposes a novel approach of getting a high resolution image from a single low resolution image. The algorithm utilizes the non sub-sampled contourlet transform to learn the coefficients of the feature representations from a dataset of images. Furthermore, the image self-similarity is presented to relieve the phenomenon of the artifact caused by the introduction of the external dictionary set. Meanwhile, we introduce the anchored neighborhood regression to solve the coefficient of sparse representation for the further improvement of efficiency of the algorithm. Experimental results show that the validity of the proposed approach through simulation on several images and measurement indicators. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|