Improved Soft Shadow Algorithm Based On First-order Distance Field

As one of the important rendering techniques,shadow technology plays a crucial role in helping people understand the geometry,size,and relative position of objects in 3D scene.The signed distance field is widely used for rendering high-quality soft shadow effects because of its ability to express 3D scene information.However,existing algorithms for soft shadow rendering using signed distance fields cause problems such as penumbra shrinking due to inaccurate estimation of occlusion rate.To solve this problem,on the basis of fully studying the characteristics of the first-order signed distance field,this thesis proposes an improved soft shadow algorithm based on the first-order signed distance field.The basic idea is that through the first-order signed distance field,the complex occlusion area of the shading point projected to the spherical light source is approximated to the union of multiple arcuate areas,so that it can analytically judge whether the sampled multiple shadow rays intersect with the arcuate areas.Combined with the Monte Carlo method,the occlusion rate can be more accurately estimated.On the basis of meeting the real-time performance,this algorithm can render higher quality soft shadow effect.Specifically,the main contribution of the work in this thesis includes:?This thesis proposes a method for estimating complex occlusion area.Aiming at the time-consuming problem of intersection calculation between shadow ray and complex occlusion area,this thesis uses multiple arch area to approximate complex occlusion area to realize the analytic intersection of shadow ray and complex occlusion area,and quickly determines the location and size of multiple arch area through the first-order partial derivative information contained in the first-order signed distance field,so that the approximation error is as small as possible.?This thesis proposes an occlusion rate calculation strategy based on arch region.According to the number of shading points occluded by arch area,different calculation methods of occlusion rate are selected.When the approximate region is a single arcuate region,the arcuate region area is calculated analytically,and then the occlusion rate is calculated quickly.However,when the approximate region is a multiple arch region,which is often overlapped,so it is difficult to solve the union area analytically.In order to solve this problem,this thesis proposes to emit multiple random shadow rays at the shading points,and then use the Monte Carlo method to accumulate the visibility of rays to achieve the fast and accurate calculation of occlusion rate.?This thesis proposes a model culling strategy based on geometric occlusion space.In the process of calculating the occlusion rate,the occlusion area of each model needs to be calculated,which brings the problem of invalid calculation because some models do not cause occlusion for shading points.To address this problem,this thesis improves the drawing efficiency by combining the geometric characteristics of surface light sources and using the geometric occlusion space formed by them and shading points to cull models that do not intersect with this space.Experimental results show that this algorithm can estimate the complex occlusion area more accurately;based on the estimated occlusion area,this algorithm can calculate the occlusion rate more accurately than the contrast algorithm,and avoid the problems of penumbra shrinking caused by inaccurate occlusion rate estimation in soft shadow algorithm based on signed distance field,which makes the soft shadow transition effect more natural.Finally,this algorithm can render more realistic soft shadow effect on the basis of real-time performance.