According to the characteristics of interference fringe region and compared with many conventional edge detection methods, a judgment function is proposed for the automatic extraction of speckle interference fringe regions based on second-order gradient entropy functions. By analyzing different speckle interference fringe images, an optimum entropy size interval of a test sub-region and adaptive threshold intervals of different interference fringe patterns are determined. The automatic extraction of the stripe region is finally completed through connected region segmentation. The method is validated experimentally. Experimental results show that more accurate image entropy is obtained when the size of sub-region entropy is above 15 pixel. The fringe region and speckle background region can be segmented accurately when the adaptive threshold is within a range of 0 to Q_max-1.25, where Q_max is the maximum gradient entropy in the image, thereby achieving automatic extraction of the speckle interference fringe region.