Detail enhancement and noise suppression are particularly important in infrared image processing, and the focus is to compress the infrared from the high to the low dynamic range while preserving details and suppressing noise. In this study, with an improved adaptive detail enhancement algorithm for infrared images based on guided filtering (AGF&DDE) as the foundation, an improved algorithm for adaptive infrared image enhancement based on guided filtering was proposed. The input image was smoothed by a guided image filter and separated into a base layer image and a detail layer image that contained large and small dynamic temperature information, respectively. Then, the base layer image and the detail layer image were compressed, noise-suppressed and finally fused with different fusion ratios to form the output image. To shorten the algorithm operation time, we highlighted the image detail information while suppressing the noise of the detail layer. The adaptive threshold parameters that can be used to screen effective gray values through the maximum and minimum values of histogram distribution information were determined, and used together with histogram distribution to design a one-dimensional compression array. The grayscale value of each pixel in the 16 bit base layer image was mapped to a range of 8bit according to the array. The effective number of grey value in the histogram was estimated, and the image scene information was judged by the ratio between the number of effective gray values to the total number of gray values. According to the different scenarios, The adaptive fusion ratio was determined, and the base layer image and the detail layer image were fused with different scale coefficients to form an 8bit output image. The experimental results were compared with the histogram equalization algorithm, the high-dynamic infrared image enhancement algorithm based on guided filtering, and the adaptive infrared image enhancement algorithm based on guided filtering. Four different scenarios were selected for analysis from subjective and objective levels. The comparison showed the image processed by this algorithm could highlight the detail contour information, reduce the influence of detail noise on the fused output image, and present a better visual effect. According to the objective evaluation, the average computing time of this algorithm in four scenarios was 0.753 5 s, which was lower compared with other algorithms. Moreover, the fusion ratio coefficient of the base layer image and the detail layer image achieved the effect of the adaptive scene.