Paper
Single Aerial Image De-Hazing and Its Parallel Computing
- Authors:
- Zhaohua Liu; Yuxia Yang; Jingyu Yang
- Abstract
- Aiming at the degraded image quality in bad weather, a novel haze removal method for single image based on dark channel prior is firstly introduced. Then these disadvantages of initialized method are analyzed and improved, a tolerance mechanism is introduced to deal with the dark channel prior ineffective zones, a fast edge-preserving interpolating algorithm of atmosphere transmission based on guided image filter is proposed to overcome time-consuming problem of soft matting algorithm, a Wallis filter is used to improve the intensity of result image. And to improve the de-hazing efficiency more, a GPU-CPU cooperating scheme is designed and corresponding fine granularity parallel computing of de-hazing is realized based on integral image and box filter through GPU. At last, three aerial images are processed using GeForce 9500 GT card to validate the proposed method.
- Keywords
- Dark Channel Prior; Guided Filter; Wallis Filter; GPU; Integral Image
- StartPage
- 33
- EndPage
- 40
- Doi