Application Research On Bidirectional Lightcut Method

The many-light method provides a general framework for rendering various illumination effects using hundreds of thousands of virtual point lights(VPLs). Within a large set of VPLs, each VPL does not contribute equally. A general, scalable algorithm tries to exploit this non-uniform VPL importance to reduce computation. It seeks to identify and evaluate the important VPLs while only sparsely evaluating the unimportant ones. To efficiently gather the contributions of these VPLs, the lightcut based methods cluster the VPLs, which implicitly approximate the lighting matrix with representative blocks. Bidirectional Lightcuts extends the previous approach to handle even more effects by adding additional receiver points per pixel.In this paper, we realizes the bidirectional lightcut algorithm based on Mitsuba System, which generates high-quality images of glossy and offline. It can also take into account the executive efficiency and the rendering effects. First, we perform the module of initialization about manylight method to generate the required data. For the lightcut integrator, we build the light tree by using the generated VPLs. For the multidimensional lightcut integrator, we also need to build the sensor tree after generating the VPSs. For the bidirectional light integrator, we generate the sampling pathes to extend the receiver path before building the sensor tree. Finally, we use all the tree-based data structures to realize the partition of lightcut or multidimensional lightcut to get the rendering results.At the last of the paper, we show the rendering results about our integrators and the analysis about the experiments.