Bas-relief Of 3D Scene And Image Edge Features Modeling And Image-based Style Design

As a kind of relief,bas-relief is a unique research field in computer graphics.At present,there are mainly three kinds of research work on bas-relief modeling.The first is to directly edit and render with 3D design software,the second is to compress the depth value based on the 3D scene,and the third is to extract its depth information based on 2D image.However,the existing bas-relief modeling algorithms based on 3D scenes relies on the control of parameters and lack of consideration of the details of the edge of bas-relief;The bas-relief produced based on two-dimensional images not only have a single style,but also lack a sense of hierarchy and stereoscopic effect.Aiming at the appeal problems,this paper starts with two research objects,3D scene and digital image,and discusses the generation scheme of bas-relief without parameter control and the method of stylized bas-relief.The specific research work is as follows:(1)In order to solve the problem that most of the bas-relief modeling algorithms rely heavily on parameters,this paper proposes a bas-relief generation scheme combining image edge features without parameter control.First obtain the depth map from the 3D scene,as a special image,depth map has low contrast.In this paper,histogram equalization algorithm is used to improve the contrast of depth map.In the stage of edge detection,considering the characteristics of depth map and the advantages of local image structure,this paper uses a random structure forest to learn the edge detector,and the depth map gets the contour of bas-relief through edge detection.Then the contour of the bas-relief is regarded as a probability matrix,and a non-linear function is constructed according to the probability value.After the gradient domain of the depth field is obtained,the function is used to carry out adaptive compression in the directional gradient domain,and then the isotropic filter is used to realize detail enhancement.Finally,the height field is restored by solving the Poisson equation,and the bas-relief model is generated.Experimental results show that the algorithm can automatically generate a bas-relief model with good effect for most types of 3D scenes in a short time without the control of artificial input parameters.(2)Aiming at the lack of stylized design in the current bas-relief modeling algorithm,this paper proposes two methods of stylized low relief based on the above image edge detection algorithm,starting from the two-dimensional image.On the one hand,only high-frequency characteristics are considered.First,binarize the edge map of the two-dimensional image,then use the mathematical morphology algorithm to denoise and connect the broken edges to obtain a complete and thickened edge contour,and finally the bas-relief model of contour style is generated.This algorithm is suitable for traditional calligraphy font and line-drawing image.On the other hand,this paper uses the relationship between image feature separation and bas-relief geometric feature,combined with the Gamma correction method and Weighted Least Squares algorithm(WLS for short)to propose two styles of bas-relief enhancement:detail enhancement and smoothing,and achieve multi-style bas-relief by linear fusion technology.It has been verified that the algorithm has good universality to most images.The above two algorithms solve the problem of single style of bas-relief generated by two-dimensional images,and have practical significance for ordinary users to obtain multi-style bas-relief.