Search

In the early modern period, many authors held that sensation or sensory reception is in some way passive and that perception is in some way active. The notion of a more passive and a more active aspect of perception is already present in Aristotle: the senses receive forms without matter more or less passively, but the ˵primary sense″ also recognizes the salience of present objects. Ibn al-Haytham distinguished ˵pure sensation″ (of light and color) from other aspects sense perception, achieved by ˵discernment, inference and recognition,″ which included perception of properties such as size and distance as well as similarity, difference, and beauty. Descartes regarded light and color as experiences passively caused in the mind by bodily processes, but he also included distance, perceived through accommodation and convergence, as an immediately caused (hence, in that way, passive) sensory idea. On the perception side, most theorists held that size and distance perception frequently occurs through unnoticed psychological operations, whether mediated by judgment (Descartes, Kant) or associative processes (Berkeley, Hume, and Reid). Association is, in a sense, passive, as it occurs through nonreflective habit formation. But such habits mark a contribution of the subject to perception and are in that way active. The decision of whether sensation and perception are active or passive is highly sensitive to what counts as activity and to what is included as sensation or perception. There is no simple formula, but the generalization that sensation is for the most part passive and perception for the most part active may stand as an imprecise summary of early modern thought on the topic. [ABSTRACT FROM AUTHOR]

The human visual cortex performs salient region detection, a process critical to the rapid understanding of a scene. This is performed on large arrays of locally interacting neurons that are slow to simulate sequentially. In this paper we describe and evaluate a novel, bio-inspired, cellular automata (CA) architecture for the determination of the salient regions within a scene. This parallel processing architecture is appropriate for implementation on a graphics processing unit (GPU). We compare the performance of this algorithm against that of CPU implemented salient region detectors. The CA algorithm is less subject to variation due to changing scale, viewpoint and illumination conditions. Also due to its GPU implementation, this algorithm is able to detect salient regions faster than the CPU implemented algorithms. [ABSTRACT FROM AUTHOR]

Computer vision approaches to saliency are based, among others, on uniqueness [1], local complexity [2], distinctiveness [3,4], spectral variation [5], and irregularity [6]. Saliency can also be viewed as the information in the data relative to a representation or model [7]. When a representation is built, a residual error is often minimised. The residual can be used to obtain saliency maps for solving challenging tasks of image and video processing. We introduce the notion of the resonant SVD and demonstrate that the SVD residual at the resonant spacing is selective to defects in spatially periodic surface textures and events in time-periodic videos. Examples with real-world images and videos are shown and discussed. [ABSTRACT FROM AUTHOR]

This paper proposes a bottom-up attention model based on pulsed Hebbian-based neural networks that simulate the lateral surround inhibition of neurons with similar visual features. The visual saliency can be represented in binary codes that simulate neuronal pulses in the human brain. Moreover, the model can be extended to the pulsed cosine transform that is very simple in computation. Finally, a dynamic Markov model is proposed to produce the human-like stochastic attention selection. Due to its good performance in eye fixation prediction and low computational complexity, our model can be used in real-time systems such as robot navigation and virtual human system. [ABSTRACT FROM AUTHOR]

Many models of attention have been implemented in recent years, but comparison and further development are difficult due to the lack of a common platform. We present JAMF, an open source simulation framework for drag & drop design and high-performance execution of attention models. Its building blocks are ˵Components″, functional units encapsulating specific algorithms. Simulations are created in the graphical JAMF client by connecting Components from the server΄s repository. Today it contains Components suitable for replication and extension of many major models of attention. Simulations are executed on the JAMF server by translation of model definitions into binary applications, while automatically exploiting the model΄s structure for parallel execution. By disentangling design and algorithmic implementation, the JAMF architecture combines a novel tool for rapid test and implementation of attention models with a high-performance execution engine. [ABSTRACT FROM AUTHOR]

Attention and memory are very closely related and their aim is to simplify the acquired data into an intelligent structured data set. Two main points are discussed in this paper. The first one is the presentation of a novel visual attention model for still images which includes both a bottom-up and a top-down approach. The bottom-up model is based on structures rarity within the image during the forgetting process. The top-down information uses mouse-tracking experiments to build models of a global behavior for a given kind of image. The proposed models assessment is achieved on a 91-image database. The second interesting point is that the relative importance of bottom-up and top-down attention depends on the specificity of each image. In unknown images the bottom-up influence remains very important while in specific kinds of images (like web sites) top-down attention brings the major information. [ABSTRACT FROM AUTHOR]

In prior work, we put forth a model of visual saliency motivated by information theoretic considerations [1]. In this effort we consider how this proposal extends to explain saliency in the spatiotemporal domain and further, propose a distributed representation for visual saliency comprised of localized hierarchical saliency computation. Evidence for the efficacy of the proposal in capturing aspects of human behavior is achieved via comparison with eye tracking data and a discussion of the role of neural coding in the determination of saliency suggests avenues for future research. [ABSTRACT FROM AUTHOR]

In this work, we show the capability of a new model of saliency, of reproducing remarkable psychophysical results. The model presents low computational complexity compared to other models of the state of the art. It is based in biologically plausible mechanisms: the decorrelation and the distinctiveness of local responses. Decorrelation of scales is obtained from principal component analysis of multiscale low level features. Distinctiveness is measured through the Hotelling΄s T2 statistic. The model is conceived to be used in a machine vision system, in which attention would contribute to enhance performance together with other visual functions. Experiments demonstrate the consistency with a wide variety of psychophysical phenomena, that are referenced in the visual attention modeling literature, with results that outperform other state of the art models. [ABSTRACT FROM AUTHOR]

In this paper we validate a new model of bottom-up saliency based in the decorrelation and the distinctiveness of local responses. The model is simple and light, and is based on biologically plausible mechanisms. Decorrelation is achieved by applying principal components analysis over a set of multiscale low level features. Distinctiveness is measured using the Hotelling΄s T2 statistic. The presented approach provides a suitable framework for the incorporation of top-down processes like contextual priors, but also learning and recognition. We show its capability of reproducing human fixations on an open access image dataset and we compare it with other recently proposed models of the state of the art. [ABSTRACT FROM AUTHOR]

Keypoints are high dimensional descriptors for local features of an image or an object. Keypoint extraction is the first task in various computer vision algorithms, where the keypoints are then stored in a database used as the basis for comparing images or image features. Keypoints may be based on image features extracted by feature detection operators or on a dense grid of features. Both ways produce a large number of features per image, causing both time and space performance challenges when upscaling the problem. We propose a novel framework for reducing the size of the keypoint database by learning which keypoints are beneficial for a specific application and using this knowledge to filter out a large portion of the keypoints. We demonstrate this approach on an object recognition application that uses a keypoint database. By using leave one out K nearest neighbor regression we significantly reduce the number of keypoints with relatively small reduction in performance. [ABSTRACT FROM AUTHOR]

How does knowledge of a domain influence the way in which we inspect artefacts from within that domain? Eye fixation scanpaths were recorded as trained individuals looked at images from within their own domain or from another domain. Sequences of fixations indicated differences in the inspection patterns of the two groups, with knowledge reflected in lower reliance of low-level visual features. Scanpaths were observed during first and second viewings of pictures and found to be reliably similar, and this relationship held in a second experiment when the second viewing was performed one week later. Eye fixation scanpaths indicate the viewer΄s knowledge of the domain of study. [ABSTRACT FROM AUTHOR]

We detail an approach to the autonomous acquisition of hierarchical perception-action competences in which capabilities are bootstrapped using an information-based saliency measure. Our principle aim is hence to accelerate learning in embodied autonomous agents by aggregating novel motor capabilities and their corresponding perceptual representations using a subsumption-based strategy. The method seeks to allocate affordance parameterizations according to the current (possibly autonomously-determined) learning goal in a manner that eliminates redundant percept-motor context, thereby obtaining maximal parametric efficiency. Experimental results within a simulated environment indicate that doing so reduces the complexity of a multistage perception-action learning problem by several orders of magnitude. [ABSTRACT FROM AUTHOR]

A novel brain inspired cognitive system architecture of an active visual search model is presented. The model is multi-modular consisting of spatial and object visual processing, attention, reinforcement learning, motor plan and motor execution modules. The novelty of the model lies on its decision making mechanisms. In contrast to previous models, decisions are made from the interplay of a winner-take-all mechanism in the spatial, object and motor salient maps between the resonated by top-down attention and bottom-up visual feature extraction and salient map formation selectively tuned by a reinforcement signal spatial, object and motor representations, and a reset mechanism due to inhibitory feedback input from the motor execution module to all other modules. The reset mechanism due to feedback inhibitory signals from the motor execution module to all other modules suppresses the last attended location from the saliency map and allows for the next gaze to be executed. [ABSTRACT FROM AUTHOR]

When searching for a target in a visual scene filled with distractors, the mechanism of inhibition of return prevents revisiting previously attended locations. We proposed a new computational model for the inhibition of return, which is able to examine priority or saliency map in a manner consistent with psychophysical findings. The basic elements of the model are two neural integrators connected with two inhibitory interneurons. The integrators keep the saliency value of the currently attended location in the working memory. The inhibitory inter-neurons modulate a feedforward flow of information between the saliency map and the output map which points to the location of interest. Computer simulations showed that the model is able to read-out the saliency map when the objects are moving or when eye movements are present. Also, it is able to simultaneously select more then one location, even when they are non-contiguous. The model can be considered as a neural implementation of the episodic theory of attention. [ABSTRACT FROM AUTHOR]

We propose an implementation of covert attention mechanisms with spiking neurons. Spiking neural models describe the activity of a neuron with precise spike-timing rather than firing rate. We investigate the interests offered by such a temporal code for low-level vision and early attentional process. This paper describes a spiking neural network which achieves saliency extraction and stable attentional focus of a moving stimulus. Experimental results obtained using real visual scene illustrate the robustness and the quickness of this approach. [ABSTRACT FROM AUTHOR]