No abstract

No abstract

In this paper, we propose a state-of-the-art video denoising algorithm based on a convolutional neural network architecture. Previous neural network based approaches to video denoising have been unsuccessful as their performance cannot compete with the performance of patch-based methods. However, our approach outperforms other patch-based competitors with significantly lower computing times. In contrast to other existing neural network denoisers, our algorithm exhibits several desirable properties such as a small memory footprint, and the ability to handle a wide range of noise levels with a single network model. The combination between its denoising performance and lower computational load makes this algorithm attractive for practical denoising applications. We compare our method with different state-of-art algorithms, both visually and with respect to objective quality metrics. The experiments show that our algorithm compares favorably to other state-of-art methods. Video examples, code and models are publicly available at https://github.com/ m-tassano/dvdnet.

The aim of motion detection is to decide whether a given part of an image belongs to a moving object or to the static background. This paper proposes an automatic decision rule for the detection of moving regions. The proposed framework is derived from a perceptual grouping principle, namely the Helmholtz principle. This principle basically states that perceptually relevant events are perceived because they deviate from a model of complete randomness. Detections are then said to be performed a contrario: moving regions appear as low probability events in a model corresponding to the absence of moving objects in the scene. A careful design of the events considered under the hypothesis of absence of moving objects results in a general and robust motion detection algorithm. No posterior parameter tuning is necessary. Furthermore, a confidence level is attached to each detected region.