Abstract:In order to improve the effect of infrared image dehazing,sky region recognition of DCP is proposed.Firstly,image hazing was divided into sky region,non sky region and cross region by double thresholds.Secondly,the atmospheric light value and transmittance of sky region,non sky region and cross region were calculated using different methods to avoid the limitations of dark channel prior.Finally,the process was given.Experimental results show that sky region recognition and dark channel prior algorithm is clear for image dehazing in the sky area,and the average gradient and comprehensive evaluation are better than other algorithms.The average gradient of the infrared image with complex cross areas between the sky area and non sky area is 28.02% higher than dark channel prior algorithm,and the comprehensive evaluation value is 19.88% higher than dark channel prior algorithm.