Comparative study of retrospective methods to reduce non-uniform illumination effects to bridge coating

Digital image processing methods have been actively studied and applied to various infrastructure components including bridge painting assessment. Despite the advantages of image recognition techniques to bridge coating, the results can be greatly affected by the non-uniform illumination of target conditions. It is necessary to first convert all images to a standardized background illumination condition. The effect caused from non-uniform illumination can be minimized by setting lighting conditions properly or capturing additional calibration images. However, capturing additional calibration images is not possible in many practical situations and hence a good method of shading correction is required. This paper addresses the way to correct digital bridge coating images containing non-uniform illumination, which becomes more difficult when it comes to assessing facility surfaces with rust defects. In order to choose the most appropriate one, four representative shading correction methods were selected and tested to identify their suitability to processing bridge coating images. These methods include: (1) image smoothing, (2) homomorphic filtering, (3) morphological filtering, and (4) entropy minimization method. Test results on bridge coating images showed that the entropy minimization method outperformed the other three methods. ► Studying how to correct bridge coating images with non-uniform illumination. ► Comparison of the various retrospective illumination correction techniques. ► Entropy minimization method proved to work best. ► All the methods are implemented in MATLAB and results are shown for comparison.