Real Time Image Haze Removal on Multi-core DSP

The quality and visual effect of an image is high demanded in avionic embedded field. But the frog and haze is common in the nature environment, so foggy images gathered in bad weather need to be disposed to remove the haze. The single image haze removal algorithm using dark channel prior can achieve great haze removal effect, but the process of optimizing the medium transmission in this algorithm costs too much time, while the computational complexity is too high to be real-time operating for high resolution image. In this paper, a novel method is proposed which uses a new kind of filter called guided filter to optimize the medium transmission. This method is much faster and also can achieve good haze removal effect. In addition, the method uses down sampling and interpolation method to transform a high resolution image into a low one to reduce the quantity of calculation. At last, we implement the novel algorithm on a multi-core DSP of TI company. The experimental results show that the method costs less than 40ms for a 600*400 resolution image and can satisfy the demand of real-time image process.