The invention provides an infrared and visible light image fusion method based on combined sparsity and nonnegative sparsity, wherein the method relates to the field of image fusion technology. A sample matrix is obtained from the infrared image and the visible image, and a fused mean value result is obtained. Combined sparse representation is performed on a mean value eliminated sample for obtaining a fused mean value eliminated result. A fused mean value result is combined with the mean value eliminated result. Significance detection is performed on the infrared image for obtaining a substantial graph, and a target area graph of the infrared image is obtained. Namely an infrared and visible light image fusion result with enhanced target can be obtained. According to the method of the invention, based on the fusion method based on combined sparse representation, a classified fusion concept based on nonnegative sparse representation is introduced. A common characteristic and a specificcharacteristic can be respectively extracted. The specific characteristic in a source image is totally reserved. The type of the characteristic can be effectively indicated. A classified fusion ruleis applied to the mean value. The details of an infrared target and background are reserved more effectively. Through comprehensive utilization of the classified fusion concept and the classified fusion rule, a better fusion result is obtained.