It is hard to quantitate the micron-scale defects on large aperture (10(2)mmx10(2)mm) optical components by the conventional optical testing methods. This paper proposes a super-smooth surface defects measurement and evaluation system, achieved by using microscopic dark-field scattering imaging device, two-dimensional sub-image scanning mechanism and multi-cycle image mosaic algorithm. The defects detecting system, with a lateral resolution of 0.5 mu m, applies a large field of view design (largest FOV: 15mmx15mm). In order to test the largest element (430mmx430mm), however, over 1000 sub-pictures are captured. It takes more than 30 minutes to process these sub-pictures by multi-cycle image mosaic algorithm. This paper also presents a distortion correction method to revise the image mosaic mismatch caused by the optical distortion in the defects testing system on the platform of MATLAB. A binary optical grid plate (BOE) is fabricated as standard board to evaluate distortion. The proposed method applies image division multi-steps to build a look-up matrix of distortion parameters. According to the look-up matrix, all pixels on a sub-image are repositioned from the distortion Cartesian coordinates to the ideal Cartesian coordinates. Finally, feasibility of the distortion correction method is demonstrated by comparing the mosaic results of defect images before and after this process. The full field view distortion is reduced from more than 4% to less than 0.1%. After distortion correction, sub-images can be directly mosaicked without using multi-cycle image mosaic algorithm, which improves test efficiency significantly. The method mentioned in this paper may also apply to other optical testing systems for image mosaic.
