Indirect Light Field Adaptive Partition And Light Field Matrix Completion Algorithm

With the rise of real-time ray tracing technology,the boundaries between real-time rendering and realistic rendering are getting blurred.Real-time ray tracing,which is anounced to be the future of graphic rendering,is still limited by the rays per pixel in the current starting stage.With the limitation of GPU performance,rendering a relatively convergent image result without obvious noise has once again become a subject of research value.Our paper's indirect light field partition and matrix completion algorithm can effectively re-store the whole light field energy distribution with a small amount of rays,meanwhile,accelerating the rendering process.The light field energy distribution conforms to the locality principle,both temporal and spatial.The algorithm in our paper makes full use of the spatial locality of the indirect light field.The energy distribution of scene's local area is sparse,and we represent its local indirect light field in the form of a low-rank matrix.Combined with the low rank matrix completion technique that mathematicians have been studying for decades,our algorithm completes the whole indirect light field matrix with limited sample rays.In addition to studying the problem of sparse light field matrix completion,our paper also deals with the problem of the non-sparse ones.That is to say,for non-diffuse materials' light field recovering,our paper proposes an adaptive light field partition method to solve the problem.At the same time,in order to reduce the negative impact of the noise inherent in the sampling process of matrix completion,our paper uses matrix approximation as an auxiliary method,filtering the matrix while completing it.Above all,our paper proposes a practical method for indirect light field adaptive partition and light field matrix completion.