Fixed pattern noise(FPN) is a kind of common noise in complementary metal oxide semiconductor(CMOS) camera sensors, which has an adverse impact on the image quality of the camera. In order to improve the quality of road disease detection images, this paper proposes a fixed-mode noise removal algorithm of CMOS image sensor based on combined filtering. Firstly, the mean filtering is used to extract the spatial distribution characteristics of the fixed-mode noise of the CMOS sensor camera. Secondly, the proposed linear filtering method is used to remove most of the fixed mode noise. Finally, nonlinear filtering is used to remove salt and pepper noise. Experimental results show that the peak signal-to-noise ratio (PSNR) of the three groups of test images is between 28.75 and 32.50, and the mean square error of the images decreases from 9.66 to 11.28 before processing to 5.45 to 7.59 after processing. The mean square error of the processed images was reduced by 27.68% to 43.61%. The proposed algorithm can effectively reduce the influence of fixed pattern noise in CMOS sensor camera on image quality and retain the main texture features of road damage images, which has higher practicability compared with the traditional calibration method for CMOS sensor camera.