In order to improve the quality of oil painting image, an image enhancement technique based on semi-supervised learning to improve the quality of oil painting image is proposed on the basis of extracting the feature points of surface light field. Moreover, a micro-lens array is used to amplify the depth difference of surface light field and a multi-angle surface light field generation and acquisition system is designed for oil painting similarity detection. In addition, in order to improve the accuracy and robustness of the quantitative calculation of the light field difference on the oil painting surface, this paper proposes the characteristic points of the light field angle domain change, and measures the similarity of the surface light field by the spatial distribution of the characteristic points. According to the experimental study, the image enhancement technology based on semi-supervised learning proposed in this paper has some advantages over the classical image enhancement technology in oil painting image enhancement.