We present an importance sampling method for the bidirectional scattering distribution function (bsdf) of hair. Our method is based on the multi‐lobe hair scattering model presented by Sadeghi et al. [SPJT10]. We reduce noise by drawing samples from a distribution that approximates the bsdf well. Our algorithm is efficient and easy to implement, since the sampling process requires only the evaluation of a few analytic functions, with no significant memory overhead or need for precomputation. We tested our method in a research raytracer and a production renderer based on micropolygon rasterization. We show significant improvements for rendering direct illumination using multiple importance sampling and for rendering indirect illumination using path tracing.

Historically, rendering system development has been mainly focused on improving the numerical accuracy of the rendering algorithms and their runtime efficiency. In this paper, we propose a method to improve the correctness not of the algorithms themselves, but of their implementation. Specifically, we show that by combining static type checking and generic programming, rendering system and shader development can take advantage of compile-time checking to perform dimensional analysis, i.e. to enforce the correctness of physical dimensions and units in light transport, and geometric space analysis, i.e. to ensure that geometric computations respect the spaces in which points, vectors and normals were defined. We demonstrate our methods by implementing a CPU path tracer and a GPU renderer which previews direct illumination. While we build on prior work to develop our implementations, the main contribution of our work is to show that dimensional analysis and geometric space checking can be successfully integrated into the development of rendering systems and shaders.

When imaging through water surface, the random fluctuation of sea surface will cause the distortion of the target scene image, so the distorted image needs to be corrected and reconstructed. At present, distortion compensation mainly adopts iterative registration strategy based on image sequences which is difficult to satisfy the real-time observation. This paper presents a correction method based on active imaging of structured light for underwater image. Experimental results show that compared with the traditional iterative algorithm, the proposed algorithm cannot only improve the restoration accuracy, but also greatly shorten the processing time. Experimental test results demonstrate that the proposed algorithm has good recovery results.

Figure 1: The light transport matrices of two complex scenes (subsampled from the original). Note the existence of repeating patterns and large areas of near black in the matrices. Our algorithm, LightSlice, effectively exploits these typical structures of light transport matrices, and efficiently solves the many-lights problem by seeking locally optimized light clustering for each slice of the light trasnport matrix. AbstractRecent work has shown that complex lighting effects can be well approximated by gathering the contribution of hundreds of thousands of virtual point lights (VPLs). This final gathering step is known as the many-lights problem. Due to the large number of VPLs, computing all the VPLs' contribution is not feasible. This paper presents LightSlice, an algorithm that efficiently solves the many-lights problem for large environments with complex lighting. As in prior work, we derive our algorithm from a matrix formulation of the many-lights problem, where the contribution of each VPL corresponds to a column, and computing the final image amounts to computing the sum of all matrix columns. We make the observation that if we cluster similar surface samples together, the slice of the matrix corresponding to these surface samples has significantly lower rank than the original matrix. We exploit this observation by deriving a two-step algorithm where we first globally cluster all lights, to capture the global structure of the matrix, and then locally refine these clusters to determine the most important lights for each slice. We then reconstruct a final image from only these locally-important lights. Compared to prior work, our algorithm has the advantage of discovering and exploiting the global as well as local matrix structure, giving us a speedup of between three and six times compared to state-of-the-art algorithms.