Aiming at the problem that the detection of weak and small infrared targets could be affected by false alarm, according to the noise characteristics of infrared line array sensor and the gradient symmetry of small target, a method of constructing multi-scale stacked enhanced re-integrated image (MSERI) in image gradient space is proposed to detect small infrared target. First, different sizes of small targets are estimated, and the unidirectional gradient images of the original image are calculated from multiple directions. Then, the complementary gradient of the gradient values in the unidirectional gradient image are found to enhance the gradient image. Then, the enhanced gradient image is integrated to restore the image, and the integrated images in different directions are stacked as a image. And, the maximum enhanced value is taken from the stacked images with different estimated sizes as the enhanced result. Finally, according to the peak to peak value of clutter in the neighborhood of pixels from enhanced result, the adaptive threshold is calculated to segment small infrared targets. The experimental results show that the method has better detection ability and lower false alarm rate in various scenarios, and the running speed is better than algorithms with similar performance.