A structured light image quality evaluation method based on no-reference quality assessment

Abstract

Affected by noise, color, and light blocking, there are some large error data contained in the results of structured light 3D measurement. In the absence of effective criterion, the data can only be recognized artificially. In this paper, a criterion based on Human Vision System (HVS), which can be applied to recognize the large error data by evaluating the noise variance level of the stripe images, and a method used to improve the level of measurement accuracy are proposed. According to the proposed method, the no-reference quality assessment is used to evaluate the quality of stripe images. With the noise evaluation coefficient, the large error data can be identified automatically. The experimental results show that this method is efficient in detecting large error data in the stripe images.