Abstract | ||
---|---|---|
Image recolorization is a process of generating new synthetic images for given reference images and input images. Color characteristics and geometrical structure details of the synthetic images are transferred from reference images and input images, respectively. In classical image recolorization models, the total variation (TV) regularizer is usually used to suppress noise and graininess during the process of estimating synthetic images. However, the TV regularizer usually produces pseudo contours known as staircase-like artifacts and cannot preserve some important structure details well. To solve these problems, in this paper a new smoothing based image recolorization model is proposed, in which a fractional-order TV regularizer is designed. Moreover, an edge protection process is also proposed which can further improve the preservation performance of image tiny details. Numerical results demonstrate that our proposed model can effectively protect image tiny details of image recolorization results, while the staircase-like artifacts are avoided.
|
Year | DOI | Venue |
---|---|---|
2019 | 10.1145/3364836.3364853 | Proceedings of the Third International Symposium on Image Computing and Digital Medicine |
Keywords | Field | DocType |
Image recolorization, augmented Lagrangian method, fractional-order total variation | Computer vision,Computer science,Smoothing,Augmented Lagrangian method,Artificial intelligence | Conference |
ISBN | Citations | PageRank |
978-1-4503-7262-6 | 0 | 0.34 |
References | Authors | |
0 | 3 |