Abstarct: Images can get various destructions that one of them is haze. Haze is formed due to the aerosols. These particles scatter the emitted light from a source and reflected light from bodies that causes an incorrect realization of the environment. Hence image dehazing is necessary. “Dark Channel” is one of the most suitable methods for dehazing. In this method, hazy image, is divided into windows. The least number of pixels is acquired from the three channels of R, G and B in a colored image in every window. Then, the value of dark pixel is obtained and replaces the whole pixels of the windows by calculating the minimum of the three foregoing values. By resulting image which is named dark channel image, haze in different sections of image is measured. Then dehazing is done according to dark channel image and airlight. This method has some problems such as producing blocking effect in smooth areas and not being useful for dehazig an image with uneven haze. In this thesis, two solutions are proposed in order to solve the problems of dark channel method. In the first solution, the smooth and no non-smooth areas are separated using image texture. Smooth regions are dehazed by Gamma correction. The dark channel image is extracted in order to dehaze non-smooth areas. Any part of the image that has a thicker haze, the corresponding dark channel image is brighter. This is why dark channel image is divided into two groups of bright and dark and dark channel method is applied again to the brighter sections. This solution shows better results comparing to other methods. The second proposed solution is an improved version of the first solution. At first, the bad effect of dark channel method on smooth regions is reduced in dark channel method. Then, the dark channel image is extracted and divided into two parts of bright and dark. This algorithm is applied on the brighter parts repeatedly. To evaluate proposed methods of qualitative and quantitative criteria are used which results demonstrate the preference of the proposed methods comparing to recent methods in the field of dark channel.