Visual Optimization Technology Based On Dark Channel Prior

In recent years, the frequency of fog and haze occurs is increasing and the outdoor visual equipments always become invalid under the influence of the fog. The images that collected by these outdoor visual equipments have problems of low contrast and fuzz, So it’s meaningful to do some research in image defogging.Three problems in the processing of defogging base on dark channel prior are proposed in this paper. Firstly, In order to suppress the noise and halo ring in the sky region, the minimum value of the transmittance is limited by the fixed threshold in the processing of the defog. Although, this method has a good effect in suppressing noise and halo ring, But the limited amount of transmittance works on the whole image space and do a bad influence in the area of heavy fog; Then, Determining the dark channel value is zero will increase the intensity of defogging and increase the contrast stretching what will cause the intensity of some pixels close to zero after defogging and make our eyes can not distinguish; Last, the third problem is the law of dark channel prior is invalid in the area where has white object. The algorithm based on dark channel prior uses fixed window to find dark channel value, if the size of white object is bigger than the window size, the law of the dark channel prior is failure.Due to the calculating of dark channel value is finding the minimum value in local region, the value we get on the edge of the white object is right and the value we get in the center of the white object is wrong, which will cause the fake edge occurs.In view of the above problems, this paper does the following improvement of the algorithm based on dark channel prior: Firstly, using the membership of the sky to distinguish between the sky and the heavy fog region, and do different adjustment according to the different membership degree. Thus, weakening the effects on heavy fog area while defogging with fixed minimum threshold. The membership coefficient of sky is changing slowly with the pixel position varies, and it does not need binarization, so the adjustment of different positions is changing slowly in the airspace, even if the detection is error, the impact of the late defog will be acceptable; Then, The essence of the processing of defogging is to contrast the tension. In order to solve the problem that the values in some area are too low after defogging, the estimated dark channel value in local region is proposed and make the image after defogging more real by more accurate estimation of dark channel value; Last, Calculating the membership coefficient to white object, and using different window size according to the coefficient. using large window in the white objects to reduce the influence of dark channel image by white objects, and using small window to calculate dark channel value in other regions to reduce theblock effect. So, we can get more accurate dark channel image.At the same time, constraining the value after defogging between 0 and 255, and make the transmittance mare accurate and eventually got a better to fog effect.