Aiming at problem that dark channel prior is invalid for large bright objects regions in single hazy image

some hazes cannot be removed using guided filtering algorithm and time complexity is higher

a fast single image dehazing algorithm based on image fusion is proposed. On the basis of atmospheric scattering model

interval estimation is carried out on atmospheric light value. Dark channel prior is adopted to get the simple estimation of transmission. Multiscale Gauss convolution is performed to get the vague estimation of transmission by Retinex Theory. After the pixels fusion between the simple estimation and vague estimation of transmission

the exact estimation of transmission is got. Cross bilateral filtering is used for smooth processing and to refine transmissivity on bright area

and the restored image is color adjusted. And final image is obtained. Experimental results show that this algorithm not only achieve good dehazing effect

and good image color

but also effectly reduce time.