Narrow Results

A Deep Spiking Neural Network (DSNN) with Heterogeneous Regularization learning technique is proposed to build a more biologically plausible approach that evaluates the amount of noise and finds a stopping criterion for fast realistic illumination. Our contribution is to introduce a model that improves the label propagation of DSNN and is more efficient on neuromorphic hardware than a corresponding Artificial Neural Network. More specifically, we develop a biological neural model with a heterogeneous regularization technique that works similarly to a human brain and can detect noise using deep spikes without relying on mathematical metrics to extract noise features. The objective function of the proposed DSNN consists of a supervised term and an unsupervised term. The supervised term enforces the matching term between the predicted labels and the known labels. The unsupervised term enforces the smoothness of the predicted labels of the entire data samples. By learning a DSNN with the proposed objective function, we are able to develop a more powerful learning algorithm. Experiments were conducted using scenes with Global Illumination and various image distortions. The proposed model was also compared with the human visual system and other state-of-the-art models. The results show better performance and advantages in terms of efficiency, an increasingly biologically plausible network, and ease of implementation in Neuromorphic Hardware.

In this paper, we present a new Monte Carlo computational way for solving the global illumination problem whereby plenty of unbiased estimators can be employed to enrich the solutions leading to simple error control and faster estimation. Especially so, the zero variance importance sampling procedure can be exploited to calculate the global illumination optimally. Based on the new scheme, a new Monte Carlo global illumination algorithm and its importance driven version have been developed and carried out. Results, which have been obtained by rendering test scenes, show that this new framework is promising.

In this paper, we propose an approximate method for global illumination. We build an occluded estimator based on an algorithm of parallel line sweep for shadow maps. The virtual point lights are classified and a method is proposed to generate irradiance estimator (IE) by azimuthally analysis of normal. For each pixel rendering, we sample the IE around the pixel to fast approximate the percentage of irradiance to the point in the scene. The experiments demonstrate that our algorithm is an effective method of approximate global illumination.

While the radiosity method gradually established itself as a main rendering technique, there are still some fundamental problems regarding the radiosity method that remain unsolved. In this paper, we present a new radisosity approach for realistic image synthesis. Unlike the conventional radiosity method, the evaluation of light energy transfer is divided into two parts. In the first part, we establish the spatial light energy distribution of the source patch. In the second part, we subdivide the receiving surfaces accordingly and perform illumination calculation. The source patch and its receiving patches are related with each other through a global cube which acts as a light energy distributor. As no form—factors are involved, the conventional constraints on both the shooting patch and the receiving patch have been removed. Illumination coherence is exploited by performing dynamic subdivision of the receiving surface with respect to each current light source and by adaptive formation of the active light source. Accurate rendering of curved surfaces, anisotropic surfaces as well as the generation of high frequency details such as shadow edges, caustic borders and bump textures are supported efficiently. Experimental results show great potentials of the new approach.