基于深度学习的图像本征属性预测方法综述

真实世界的外观主要取决于场景内对象的几何形状、表面材质及光照的方向和强度等图像的本征属性。通过二维图像预测本征属性是计算机视觉和图形学中的经典问题，对于图像三维重建、增强现实等应用具有重要意义。然而二维图像的本征属性预测是一个高维的、不适定的逆向问题，通过传统算法无法得到理想结果。针对近年来随着深度学习在二维图像处理各个方面的应用，出现的大量利用深度学习对图像本征属性进行预测的研究成果，首先介绍了基于深度学习的图像本征属性预测算法框架，分析了以获得场景反射率和阴影图为主的本征图像预测、以获得图像中材质 BRDF 参数为主的本征属性预测及以获得图像光照相关信息为主的本征属性预测 3 个方向的国内外研究进展并总结了各自方法的优缺点，最后指出了图像本征属性预测的研究趋势和重点。

Review on deep learning based prediction of image intrinsic properties

 The appearance of the real world primarily depends on such intrinsic properties of images as the geometryof objects in the scene, the surface material, and the direction and intensity of illumination. Predicting these intrinsicproperties from two-dimensional images is a classical problem in computer vision and graphics, and is of greatimportance in three-dimensional image reconstruction and augmented reality applications. However, the prediction ofintrinsic properties of two-dimensional images is a high-dimensional and ill-posed inverse problem, and fails to yieldthe desired results with traditional algorithms. In recent years, with the application of deep learning to various aspectsof two-dimensional image processing, a large number of research results have predicted the intrinsic properties ofimages through deep learning. The algorithm framework was proposed for deep learning-based image intrinsicproperty prediction. Then, the progress of domestic and international research was analyzed in three areas: intrinsicimage prediction based on acquiring scene reflectance and shading map, intrinsic properties prediction based on acquiring material BRDF parameters, and intrinsic properties prediction based on acquiring illumination-relatedinformation. Finally, the advantages and disadvantages of each method were summarized, and the research trends andfocuses for image intrinsic property prediction were identified.