Haze is a complex atmospheric phenomenon formed by the combined effects of fog and haze that widely affects the environment, health, and economy in many parts of the world. Haze can reduce image contrast and blur details, affecting the visual effect. The existence of haze also seriously impacts our production and life. For vision-based intelligent machines, haze seriously reduces the quality of images captured by vision-based intelligent devices, affecting the next step of image processing and analysis. Image defogging is divided into defogging algorithms based on image enhancement, defogging methods based on image restoration, and defogging methods based on deep learning. The above three defogging methods have corresponding advantages and disadvantages, but the defogging effect in the bright region of the image and the model's anti-noise performance is not ideal. To ensure that the defogging model has a good defogging effect in the bright region, and at the same time has a good anti-noise performance, this paper designs an image defogging algorithm based on the transmittance multi-guidance and sharpening compensation (Fig.1).The image defogging algorithm based on transmittance multi-guidance and sharpening compensation firstly applies the method of threshold segmentation to solve the atmospheric light value, which not only sets zero the pure white pixel points in the original and dark channel images but also defines the segmentation of the white region in the original image, which solves the problem that the atmospheric light value takes the value of the bright region such as white in the original image (Fig.2); Secondly, to ensure that the proposed model can effectively deal with different areas in the image, a multiple bootstrap method is designed for transmittance taking, which converts the distortion problem of bright regions into a transmittance taking error reduction problem. In addition, Gaussian filtering is introduced into the three-channel image for noise reduction, which realizes the defogging and improves the noise resistance of the model; Finally, image sharpening is used to enhance the defogging results to improve the model's ability to recover the details at the edges. The brightness is adjusted by setting the target to complete the depth compensation of the current brightness to the target brightness, to achieve the joint optimization of the edge details of the image after defogging and the visualization effect.The SOTS dataset is firstly investigated to calculate the magnitude of the mean transmittance of bright and non-bright regions to determine the transmittance multiple bootstrap parameters (Tab.1). To verify the de-fogging effect and generalization performance of the proposed model, this paper applies the proposed algorithm and the mainstream de-fogging algorithms to conduct comparative experiments on the thin fog, medium fog, dense fog, and real haze datasets, respectively. From the experimental results under the foggy dataset, it can be seen that the results processed by the DTGSC algorithm obtain the most excellent defogging performance on the foggy dataset, and the average value of the image SSIM reaches 91.99%. It shows that the de-fogged image obtained by using the proposed model is the closest to the original image with the lowest distortion, and verifies the effectiveness of the proposed algorithm (Fig.3, Tab.2); As can be seen from the experimental results of the ablation of image sharpening and luminance compensation (Fig.4-Fig.5, Tab.3-Tab.4), after image sharpening and luminance compensation, the overall performance of the image after defogging is enhanced, which helps improve the defogging effect; as can be seen from the experimental results in a medium haze scene, the overall defogging effect obtained by the DTGSC algorithm is better in medium haze scenes (Fig.6, Tab.5), but there is some room for improvement in the image edge detail recovery, there is some room for improvement. The average SSIM value of the image obtained by the DTGSC algorithm reaches 89.80%, which has an obvious advantage over other mainstream defogging algorithms, indicating that the resultant image obtained by applying the proposed defogging algorithm has a higher degree of similarity with the original image; from the experimental results in the foggy scenario, it can be seen that the average value of the performance obtained by applying the DTGSC algorithm is still on a relatively high trend (Fig.7, Tab.6). The average PSNR performance obtained in the fog scene is 38.89 dB, and the average image SSIM performance reaches 86.29%, which also has obvious advantages over other defogging algorithms; from the experimental results under the real foggy dataset, it can be seen that the average value of the image PSNR obtained by using the DTGSC algorithm is 38.94 dB, and the average value of the image SSIM is 83.25%. 83.25%, the above results fully verify the effectiveness of the proposed defogging method (Fig.8, Tab.7). From the defogging efficiency, it is known that the DTGSC algorithm has good defogging efficiency (Tab.8).The DTGSC algorithm obtains an average image MSE value of 11.07, an average PSNR value of 39.78 dB, an average SSIM value of 87.83% under the four datasets, and an average de-fogging time on the thin fog dataset of 0.63 s. Relative to the DCMPNet algorithm, the average MSE value is scaled down by 20.54 dB, the average PSNR value is improved by 5.57 dB, the SSIM value is improved by 2.52% on average, and the de-fogging efficiency is improved by 0.08s on average. The above experimental results illustrate that the DTGSC algorithm has good performance under all three foggy datasets, verifying the effectiveness and superiority of the proposed algorithm.