A visual attention system, inspired by the behavior and the neuronal architecture of the early primate visual system, is presented. Multiscale image features are combined into a single topographical saliency map. A dynamical neural network then selects attended locations in order of decreasing saliency. The system breaks down the complex problem of scene understanding by rapidly selecting, in a computationally efficient manner, conspicuous locations to be analyzed in detail.

/pdf/a-model-of-saliency-based-visual-attention-for-rapid-scene-wv0n3yqxix.pdf

A model of saliency-based visual attention for rapid scene analysis

Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

Data Mining Practical Machine Learning Tools and Techniques

This paper presents a method for recognizing scene categories based on approximate global geometric correspondence. This technique works by partitioning the image into increasingly fine sub-regions and computing histograms of local features found inside each sub-region. The resulting "spatial pyramid" is a simple and computationally efficient extension of an orderless bag-of-features image representation, and it shows significantly improved performance on challenging scene categorization tasks. Specifically, our proposed method exceeds the state of the art on the Caltech-101 database and achieves high accuracy on a large database of fifteen natural scene categories. The spatial pyramid framework also offers insights into the success of several recently proposed image descriptions, including Torralbas "gist" and Lowes SIFT descriptors.

/pdf/beyond-bags-of-features-spatial-pyramid-matching-for-4aa5cbe2fz.pdf

Beyond Bags of Features: Spatial Pyramid Matching for Recognizing Natural Scene Categories

A multimodal system for the non-visual exploration of digital pictures

A BCI allows a person to communicate with the external world using artificial electronic or mechanical devices controlled by means of brain signals. Present-day BCIs can be divided into invasive and noninvasive. Prospective application of invasive BCIs to humans is debatable, since they require direct recording of electrical activity from the cortex with the inherent medical risks. Furthermore, the quality of the signals directly recorded on the brain deteriorates over time requiring new surgical interventions and implants in order to keep the functionality of the device. Non-invasive BCI basically rely on specific scalp EEG features. The main disadvantage of latter methods is that scalp EEG signals represent the noisy spatiotemporal overlapping of activity arising from very diverse brain regions, i.e., a single scalp electrode picks up and mixes the temporal activity of myriads of neurons at very different brain areas. Consequently, temporal and spectral features, specific to different processes arising at different areas, are intermixed on the same recording. Here we describe our approach to develop a direct noninvasive BCI system aimed to reproduce the excellent speed and prediction properties of the invasive systems while suppressing their risks. For doing that, we propose the non-invasive estimation of local field potentials in the whole human brain from the scalp measured EEG data using recently developed inverse solutions (LAURA and ELECTRA) to the neuroelectromagtic inverse problem. Recent studies have shown that the temporal and spectral features of the electric activity within the brain can be estimated from the surface signals with high precision. The goal of a linear inverse procedure is to de-convolve or un-mix the scalp signals attributing to each brain area its own temporal activity. By targeting on the particular temporal/spectral features at specific brain areas we expect to select a low number of features that capture information related to the state of the individual in a way that is relatively invariant to time. This avoids long training periods and increases the reliability and efficiency of the classifiers. For instances, in paralyzed patients the classification stage can be improved by focusing on the specific brain areas known to participate and code the different steps of voluntary or imagined motor action trough temporal and spectral features. For these reasons, a direct non invasive BCI system based on linear inverse solutions should combine the advantages of invasive and non invasive devices providing a safer and faster alternative to translate brain thoughts into actions.

Direct non-invasive brain computer interfaces

This paper introduces a content-based information retrieval method inspired by the ideas of spreading activation models. In response to a given query,the proposed approach computes document ranks as their final activation values obtained upon completion of a diffusion process. This diffusion process,in turn,is dual in the sense that it models the spreading of the query 's initial activation simultaneously in two similarity domains: low-level feature-based and high-level semantic.The formulation of the diffusion process relies on an approximation that makes it possible to compute the final activation as a solution to a linear system of differential equations via a matrix exponential without the need to resort to an iterative simulation.The latter calculation is performed efficiently by adapting a sparse routine based on Krylov sub-space projection method.The empirical performance of the described dual diffusion model has been evaluated in terms of precision and recall on the task of content-based digital image retrieval in query-by-example scenario. The obtained experimental results demonstrate that the proposed method achieves better overall performance compared to traditional feature-based approaches. This performance improvement is attained not only when both similarity domains are used, but also when a diffusion model operates only on the feature-based similarities.

/pdf/dual-diffusion-model-of-spreading-activation-for-content-7wimq6t2se.pdf

Dual diffusion model of spreading activation for content-based image retrieval

In this work, we consider the text data-hiding problem as a particular instance of the well-known Gel'fand-Pinsker problem [1]. The text, where some message $m \in \mathcal{M}$ is to be hidden, is represented by x and called cover text. Each component xi, i = 1, 2, ..., N, of x represents one character from this text. Here, we define a character as an element from a given language alphabet (e.g. the latin alphabet {A, B, ..., Z}). To be more precise, we conceive each character xi as a data structure consisting of multiple component fields (features): name, shape, position, orientation, size, color, etc.

/pdf/a-theoretical-framework-for-data-hiding-in-digital-and-4q13nn07um.pdf

A theoretical framework for data-hiding in digital and printed text documents

In this paper, the problem of capacity analysis of data-hiding techniques in a game information-theoretic framework is considered. Capacity is determined by the stochastic model of the host image, by the distortion constraints, and by the side information about the watermarking channel state available at the encoder and at the decoder. The importance of the proper modeling of image statistics is emphasized, and for this purpose, a novel stochastic nonstationary image model is proposed that is based on geometrical priors, the so-called edge process model. Being mathematically simple and tractable, the edge process model outperforms the estimation-quantization (EQ) and spike process models in reference applications such as denoising. Finally, this model allows us to obtain a realistic estimate of maximal embedding rates, and in particular, it is shown that the expected capacity limit of real images is significantly lower than previously reported.

/pdf/the-edge-process-model-and-its-application-to-information-33u57iyws7.pdf

Thierry Pun

Papers

A multimodal system for the non-visual exploration of digital pictures

Direct non-invasive brain computer interfaces

Dual diffusion model of spreading activation for content-based image retrieval

A theoretical framework for data-hiding in digital and printed text documents

The edge process model and its application to information-hiding capacity analysis