Abstract: Experiments indicate that the traditional methods of infrared image detection are not ideal because of the bigger noise of infrared image. The algorithm based on multidegree andorientation gradient fusion is presented to recognize the dotobiects in the lowcontrast infrared image polluted by heavy noise and stray waves. Experiments for detecting dim dottargets with the above method are carried out. The result indicates that the detection performance of the multidegree andorientation gradient fusion is better than that of the adaptive filter and medium filter. The algorithm is especially suited to the surveillance systems of low frame rate and can detect the dottargets with signal to clutter ratio equal to 1.