A method for evaluating the infrared signatures of decoys by using small patches and entropy is proposed. Firstly, a small patch method is used to segment the image patches with high values, then the image matching probability is calculated and the evaluation standard is determined by using entropy, and finally the effectiveness of the decoy is evaluated. The higher the matching probability between the real target and the decoy is, the greater the similarity of them is and the better the effectiveness of the decoy is. This method can eleminate the error in image information extraction very well.