Specular Detection and Removal for a Grayscale Image Based on the Markov Random Field
Specular detection and removal has been a hot topic in the field of computer vision. Most of the existing methods are mainly for color images, but grayscale images are widely used. For a single grayscale image with only intensity information, highlight de
- PDF / 1,271,073 Bytes
- 9 Pages / 439.37 x 666.14 pts Page_size
- 61 Downloads / 176 Views
1
)
School of Computer Science and Technology, Harbin University of Science and Technology, Harbin 150080, China 2 School of Computer Science and Technology, Harbin Institute of Technology, Harbin 150001, China 3 Instrument Science and Technology Postdoctoral Research Station, Harbin University of Science and Technology, Harbin 150080, China [email protected]
Abstract. Specular detection and removal has been a hot topic in the field of computer vision. Most of the existing methods are mainly for color images, but grayscale images are widely used. For a single grayscale image with only intensity information, highlight detection and removal becomes a difficult issue. To solve this problem, the single grayscale image highlight detection and removal method based on Markov random field is presented. Each reflection component modeling is estimated by geometric relation of surface normal in diffuse and specular reflection component in the framework of Markov random field. Their maximum a posteriori estimation is calculated under Bayesian formula and highlight area is detected. Finally, image inpainting method based on the BSCB model removes highlights. Experiment reveals that this method can effectively detect grayscale image specular reflection area, improve highlight areas the repair rate. Keywords: Computer vision · Specular detection · Markov random field
1
Introduction
An opaque object image is formed by reflecting incident light on the surface. Highlight presents the characteristics of the light source mainly, but it can be seen as the surface characteristics of the object in the visual effects. When the general image intensity values below a certain value, it belongs to the scope of diffuse, meanwhile, people’s vision feel softer effect. When the reflected light is very strong, the image shows a highlight effect, and then people will have dazzling visual sense. Because of the highlight, surface texture features will weaken or even disappear, the original color of the object is obscured. Highlight results in losing partial areas information and affects the image quality, which handles computer vision image is a big distraction. It often leads to image segmentation, recognition and matching error. To be able to extract accurate object feature information and ensure that the image can be applied in image segmentation, recognition and matching, and other fields, highlights detection and removal technology is essential. Most of the existing highlight detection method is mainly for color image, rarely for the © Springer Science+Business Media Singapore 2016 W. Che et al. (Eds.): ICYCSEE 2016, Part I, CCIS 623, pp. 641–649, 2016. DOI: 10.1007/978-981-10-2053-7_57
642
F. Yin et al.
grayscale image, but the grayscale image is very common in the field of computer vision. For a small amount of information available, grayscale image highlight detection and removal is a difficult problem. To solve this problem, gray image specular detection and removal method based on Markov random field is presented.
2
Related Work
There are many single
Data Loading...
