The invention discloses a dark channel-based image defogging method for linear self-adaptive improvement of global atmospheric light. The method comprises the steps of firstly, obtaining a haze imagein a hazy weather; secondly obtaining an image atmospheric light change angle theta through a method for calculating a slope of a connection line of centers of gravity for a binary image of the image;thirdly calculating a linear atmosphere light graph changed in an atmospheric light change direction theta; and fourthly, solving a fog-free image through an atmospheric scattering model, and outputting a processed foggy image in the hazy weather. According to the method, the situation that a distant scene is not distorted under the condition of dense fog or relatively high depth of field is satisfied, and the details of the distant scene are retained; and the method is particularly important for subsequent haze image processing and information extraction.