Search

Deep Neural Networks (DNNs) have been deployed in many real-world applications in various domains, both industry and academic, and have proven to deliver outstanding performance. However, DNNs are vulnerable to adversarial attacks, that are small perturbations embedded in an image. As a result, introduction of DNNs into safety-critical systems, such as autonomous vehicles, unmanned aerial vehicles or healthcare devices, would introduce very high risk of limiting their capabilities to recognize and interpret the environment in which they are used and therefore would lead to devastating consequences. Thus, robustness enhancement of DNNs by development of defense mechanisms is a matter of the utmost importance. In this paper, we evaluated a set of state-of-the-art denoising filters designed for impulsive noise removal as defensive solutions. The proposed methods are applied as a pre-processing step, in which the adversarial patterns in the source image are removed before performing classification task. As a result, the pre-processing defense block can be easily integrated with any type of classifier, without any knowledge about utilized training procedures or internal architecture of the model. Moreover, the evaluated filtering methods can be considered as universal defensive techniques, as they are completely unrelated with the internal aspects of the selected attack and can be applied against any type of adversarial threats. The experimental results obtained on German Traffic Sign Recognition Benchmark (GTSRB) have proven that the denoising filters provide high robustness against sparse adversarial attacks and do not significantly decrease the classification performance on non-altered data.

The substantial improvement in the efficiency of switching filters, intended for the removal of impulsive noise within color images is described. Numerous noisy pixel detection and replacement techniques are evaluated, where the filtering performance for color images and subsequent results are assessed using statistical reasoning. Denoising efficiency for the applied detection and interpolation techniques are assessed when the location of corrupted pixels are identified by noisy pixel detection algorithms and also in the scenario when they are already known. The results show that improvement in objective quality measures can be achieved by using more robust detection techniques, combined with novel methods of corrupted pixel restoration. A significant increase in the image denoising performance is achieved for both pixel detection and interpolation, surpassing current filtering methods especially via the application of a convolutional network. The interpolation techniques used in the image inpainting methods also significantly increased the efficiency of impulsive noise removal.

Deep learning has been widely applied in many computer vision tasks due to its impressive capability of automatic feature extraction and classification. Recently, deep neural networks have been used in image denosing, but most of the proposed approaches were designed for Gaussian noise suppression. Therefore, in this paper, we address the problem of impulsive noise reduction in color images using Denoising Convolutional Neural Networks (DnCNN). This network architecture utilizes the concept of deep residual learning and is trained to learn the residual image instead of the directly denoised one. Our preliminary results show that direct application of DnCNN allows to achieve significantly better results than the state-of-the-art filters designed for impulsive noise in color images.

Noise reduction is one of the most important and still active research topics in low-level image processing due to its high impact on object detection and scene understanding for computer vision systems. Recently, we observed a substantially increased interest in the application of deep learning algorithms. Many computer vision systems use them, due to their impressive capability of feature extraction and classification. While these methods have also been successfully applied in image denoising, significantly improving its performance, most of the proposed approaches were designed for Gaussian noise suppression. In this paper, we present a switching filtering technique intended for impulsive noise removal using deep learning. In the proposed method, the distorted pixels are detected using a deep neural network architecture and restored with the fast adaptive mean filter. The performed experiments show that the proposed approach is superior to the state-of-the-art filters designed for impulsive noise removal in color digital images.

The observations of solar photosphere from the ground encounter significant problems due to the presence of Earth's turbulent atmosphere. Prior to applying image reconstruction techniques, the frames obtained in most favorable atmospheric conditions (so-called lucky frames) have to be carefully selected. However, the estimation of the quality of images containing complex photospheric structures is not a trivial task and the standard routines applied in night-time Lucky Imaging observations are not applicable. In this paper we evaluate 36 methods dedicated for the assessment of image quality which were presented in the rich literature over last 40 years. We compare their effectiveness on simulated solar observations of both active regions and granulation patches, using reference data obtained by the Solar Optical Telescope on the Hindoe satellite. To create the images affected by a known degree of atmospheric degradation, we employ the Random Wave Vector method which faithfully models all the seeing characteristics. The results provide useful information about the methods performance depending on the average seeing conditions expressed by the ratio of the telescope's aperture to the Fried parameter, $D/r_0$. The comparison identifies three methods for consideration by observers: Helmli and Scherer's Mean, Median Filter Gradient Similarity, and Discrete Cosine Transform Energy Ratio. While the first one requires less computational effort and can be used effectively virtually in any atmospherics conditions, the second one shows its superiority at good seeing ($D/r_0, submitted to Solar Physics

We propose a novel filtering technique capable of reducing the multiplicative noise in ultrasound images that is an extension of the denoising algorithms based on the concept of digital paths. In this approach, the filter weights are calculated taking into account the similarity between pixel intensities that belongs to the local neighborhood of the processed pixel, which is called a path. The output of the filter is estimated as the weighted average of pixels connected by the paths. The way of creating paths is pivotal and determines the effectiveness and computational complexity of the proposed filtering design. Such procedure can be effective for different types of noise but fail in the presence of multiplicative noise. To increase the filtering efficiency for this type of disturbances, we introduce some improvements of the basic concept and new classes of similarity functions and finally extend our techniques to a spatiotemporal domain. The experimental results prove that the proposed algorithm provides the comparable results with the state-of-the-art techniques for multiplicative noise removal in ultrasound images and it can be applied for real-time image enhancement of video streams.

In this paper, we propose an automatic method of facial expressions recognition in static images using the high dimensional Local Binary Patterns (LBP). In this research some existing algorithms for face detection, facial landmarks localization, face normalization and recognition were combined and adopted for facial expressions classification. The novelty of the contribution lies in the application of the Random Frog algorithm used in the gene selection in the microarray experiments together with the high dimensional LBP as features of the Support Vector Machines (SVM) classifier used for the facial expressions classification. The proposed method was evaluated on the Static Facial Expressions in the Wild (SFEW) database, which contains the face images coming from movies and is very similar to real world scenarios. The performed tests show that the high dimensional LBP and features calculated from the regions around facial landmarks can achieve significant improvements over the state-of-the-art methods.

In this paper a new method of eye blink detection and analysis is proposed. The described technique is based on a combination of spatial and temporal derivatives calculated in video sequences acquired with a high speed camera. The pixels of each frame are divided into two groups according to the direction and magnitude of the hybrid gradient vectors and the distance between their centers of gravity is used for the determination of the eye movement characteristics. Preliminary experiments revealed a high detection rate of eye blinks, which is not dependent on the change of head position. The proposed technique can also be used for the analysis of eye movements and can be utilized in systems which are monitoring human fatigue, drowsiness and emotional states.