Stopping Criterion during Rendering of Computer-Generated Images Based on SVD-Entropy

Abstract

The estimation of image quality and noise perception still remains an important issue in various image processing applications. It has also become a hot topic in the field of photo-realistic computer graphics where noise is inherent in the calculation process. Unlike natural-scene images, however, a reference image is not available for computer-generated images. Thus, classic methods to assess noise quantity and stopping criterion during the rendering process are not usable. This is particularly important in the case of global illumination methods based on stochastic techniques: They provide photo-realistic images which are, however, corrupted by stochastic noise. This noise can be reduced by increasing the number of paths, as proved by Monte Carlo theory, but the problem of finding the right number of paths that are required in order to ensure that human observers cannot perceive any noise is still open. Until now, the features taking part in the human evaluation of image quality and the remaining perceived noise are not precisely known. Synthetic image generation tends to be very expensive and the produced datasets are high-dimensional datasets. In that case, finding a stopping criterion using a learning framework is a challenging task. In this paper, a new method for characterizing computational noise for computer generated images is presented. The noise is represented by the entropy of the singular value decomposition of each block composing an image. These Singular Value Decomposition (SVD)-entropy values are then used as input to a recurrent neural network architecture model in order to extract image noise and in predicting a visual convergence threshold of different parts of any image. Thus a new no-reference image quality assessment is proposed using the relation between SVD-Entropy and perceptual quality, based on a sequence of distorted images. Experiments show that the proposed method, compared with experimental psycho-visual scores, demonstrates a good consistency between these scores and stopping criterion measures that we obtain.

Heterogeneous Regularization for Fast Rendering using Deep Spike Neural Network

“…We performed our experiments on scenes with 800 × 800 resolution computed using the path tracing algorithm (figure 25). The images were cut into 16 blocks of size 200 × 200 pixels [15]. The number of paths per pixel between two consecutive images, the maximum number of paths per pixel, and the batch size were set differently for each scene to avoid evaluating noise perception models on redundant images (Table 6).…”

Guided-Generative Network for noise detection in Monte-Carlo rendering

“…Finally, the two masks are passed to an SVM model that allows to classify the current image block as still noisy or not where each block is of size 128 × 128. Giving 2 images More recently [10] proposed to use the SVD-Entropy [11] measurement as a noise feature. They compute the SVD-Entropy on each block, taking into account a sliding window of size S for which each version of the block has a decreasing noise level.…”

“…In this paper, we aim to propose an approach where these features are generated automatically before binary classification. For this purpose, we rely on the notion of noise mask [7] which will be provided by a generative neural network model and the use of a sliding window of images that appears to bring a better robustness of the models [10]. The proposal for such an approach is justified by the emergence of robust deep convolution learning methods for both noise processing and recognition [13,14].…”

“…The model obtained after a training of 30 epochs on the training dataset is compared to the RNN proposed by [10] with the same training conditions, i.e. the same train and test dataset, the same batch size and the same sliding window size (S = 6).…”

No-Reference Quality Assessment for 3D Synthesized Images Based on Visual-Entropy-Guided Multi-Layer Features Analysis

“…With the advancement of video technologies, a free viewpoint video (FVV) system is gradually applied to various fields, such as distance education, medical service, and entertainment [ 1 ]. Compared with traditional 2D videos, users can interactively embody 3D scenes from arbitrary viewpoints in the FVV system.…”