Robust visual tracking based on online learning sparse representation

Handling appearance variations is a very challenging problem for visual tracking. Existing methods usually solve this problem by relying on an effective appearance model with two features: (1) being capable of discriminating the tracked target from its background, (2) being robust to the target's appearance variations during tracking. Instead of integrating the two requirements into the appearance model, in this paper, we propose a tracking method that deals with these problems separately based on sparse representation in a particle filter framework. Each target candidate defined by a particle is linearly represented by the target and background templates with an additive representation error. Discriminating the target from its background is achieved by activating the target templates or the background templates in the linear system in a competitive manner. The target's appearance variations are directly modeled as the representation error. An online algorithm is used to learn the basis functions that sparsely span the representation error. The linear system is solved via ℓ1 minimization. The candidate with the smallest reconstruction error using the target templates is selected as the tracking result. We test the proposed approach using four sequences with heavy occlusions, large pose variations, drastic illumination changes and low foreground-background contrast. The proposed approach shows excellent performance in comparison with two latest state-of-the-art trackers.

Robust Audio-Visual Speech Recognition under Noisy Audio-Video Conditions

This paper presents the maximum weighted stream posterior (MWSP) model as a robust and efficient stream integration method for audio-visual speech recognition in environments, where the audio or video streams may be subjected to unknown and time-varying corruption. A significant advantage of MWSP is that it does not require any specific measurements of the signal in either stream to calculate appropriate stream weights during recognition, and as such it is modality-independent. This also means that MWSP complements and can be used alongside many of the other approaches that have been proposed in the literature for this problem. For evaluation we used the large XM2VTS database for speaker-independent audio-visual speech recognition. The extensive tests include both clean and corrupted utterances with corruption added in either/both the video and audio streams using a variety of types (e.g., MPEG-4 video compression) and levels of noise. The experiments show that this approach gives excellent performance in comparison to another well-known dynamic stream weighting approach and also compared to any fixed-weighted integration approach in both clean conditions or when noise is added to either stream. Furthermore, our experiments show that the MWSP approach dynamically selects suitable integration weights on a frame-by-frame basis according to the level of noise in the streams and also according to the naturally fluctuating relative reliability of the modalities even in clean conditions. The MWSP approach is shown to maintain robust recognition performance in all tested conditions, while requiring no prior knowledge about the type or level of noise.

Transpupillary thermotherapy for choroidal melanoma:Tumor control and visual results in 100 consecutive cases

Objective: The authors evaluated the results of primary transpupillary thermotherapy for choroidal melanoma in 100 cases. Design: Prospective nonrandomized analysis of treatment method. Participants: One hundred patients with choroidal melanoma were studied. Main Outcome Measures: Tumor response, ocular side effects, and visual results. Results: Of 100 consecutive patients with choroidal melanoma treated with transpupillary thermotherapy, the mean tumor basal diameter was 7.1 mm and tumor thickness was 2.8 mm. The tumor margin touched the optic disc in 34 eyes (34%) and was beneath the fovea in 42 eyes (42%). Documented growth was present in 64 eyes (64%), and known clinical risks for growth were present in all of the remaining 36 eyes (36%), with an average of 4 of 5 statistical risk factors for growth per tumor. After a mean of three treatment sessions and 14 months of follow-up, the mean tumor thickness was reduced to 1.4 mm. Treatment was successful in 94 eyes (94%) and failed in 6 eyes (6%). Three patients with amelanotic tumors showed no initial response to thermotherapy, but subsequent intravenous indocyanine green administration during thermotherapy resulted in improved heat absorption and tumor regression to a flat scar. The six eyes classified as treatment failures included four eyes with tumors that showed partial or no response to thermotherapy, thus requiring plaque radiotherapy or enucleation, and two eyes with recurrence, subsequently controlled with additional thermotherapy. After treatment, the visual acuity was the same (within 1 line) or better than the pretreatment visual acuity in 58 eyes (58%) and worse in 42 eyes (42%). The main reasons for poorer vision included treatment through the foveola for subfoveal tumor (25 eyes), retinal traction (10 eyes), retinal vascular obstruction (5 eyes), optic disc edema (1 eye), and unrelated ocular ischemia (1 eye). Temporal location (versus nasal and superior, P = 0.02) and greater distance from the optic disc (P = 0.04) were risks for retinal traction. Conclusions: Transpupillary thermotherapy may be an effective treatment for small posterior choroidal melanoma, especially those near the optic disc and fovea. Despite satisfactory local tumor control, ocular side effects can result in decreased vision. Longer follow- up will be necessary to assess the impact of thermotherapy on ultimate local tumor control and metastatic disease.

Visual outcomes in the subfoveal radiotherapy study:A randomized controlled trial of teletherapy for age-related macular degeneration

Objective: To determine whether teletherapy with 6-mV photons can reduce visual loss in patients with subfoveal choroidal neovascularization in age-related macular degeneration. Design: A multicenter, single-masked, randomized controlled trial of 12 Gy of external beam radiation therapy delivered to the macula of an affected eye vs observation only. Setting: Three United Kingdom-based hospital units. Participants: Patients with age-related macular degeneration, aged 60 years and older, who had subfoveal choroidal neovascularization and a visual acuity of 20/200 (logMAR 1.0) or better. Methods: Two hundred three patients were randomly assigned to radiotherapy or observation. Treatment was undertaken at designated radiotherapy centers, and patients assigned to the treatment group received a total dosage of 12 Gy of 6-mV photons in 6 fractions. Follow-up was scheduled at 3, 6, 12, and 24 months. After excluding protocol violators, the data from 199 patients were analyzed. Main Outcome Measures: The primary outcome measure was mean loss of distance visual acuity in the study eye at 12 and 24 months. Other outcome variables analyzed were near visual acuity and contrast sensitivity. The proportions of patients losing 3 or more or 6 or more lines of distance and near acuity and 0.3 or more or 0.6 or more log units of contrast sensitivity at each follow-up were also analyzed. Results: At all time points, mean distance visual acuity was better in the radiotherapy-treated group than in the control group, but the differences did not reach statistical significance. At 24 months, analysis of the proportions of patients with loss of 3 or more (moderate) (P=.08) or 6 or more (severe) (P=.29) lines of distance vision showed that fewer treated patients had severe losses, but there was no statistically significant difference between groups. For near visual acuity, although there was no evidence of treatment benefit at 12 and 24 months, a significant difference in favor of treatment was present at 6 months (P=.048). When analyzed by the proportions of patients losing 3 lines of contrast sensitivity, there was a significant difference in favor of treatment at 24 months (P=.02). No adverse retinal effects were observed during the study, but transient disturbance of the precorneal tear film was noted in treated patients. Conclusion: The results of the present trial do not support the routine clinical use of external beam radiation therapy in subjects with subfoveal choroidal neovascularization in age-related macular degeneration.

What is the best method for diagnosing glaucoma?

Glaucoma is characterized by a typical appearance of the optic disc and peripheral visual field loss. However, diagnosis may be challenging even for an experienced clinician due to wide variability among normal and glaucomatous eyes. Standard automated perimetry is routinely used to establish the diagnosis of glaucoma. However, there is evidence that substantial retinal ganglion cell damage may occur in glaucoma before visual field defects are seen. The introduction of newer imaging devices such as confocal scanning laser ophthalmoscopy, scanning laser polarimetry and optical coherence tomography for measuring structural changes in the optic nerve head and retinal nerve fiber layer seems promising for early detection of glaucoma. New functional tests may also help in the diagnosis. However, there is no evidence that a single measurement is superior to the others and a combination of tests may be needed for detecting early damage in glaucoma. © 2010 Expert Reviews Ltd.

Comparison of two fast strategies, SITA fast and TOP, for the assessment of visual fields in glaucoma patients

Purpose: To compare two fast threshold strategies of visual field assessment; SITA Fast (HSF) and Tendency Orientated Perimetry (TOP), in detecting visual field loss in patients with glaucoma. Methods: Seventy-six glaucoma, ocular hypertensive and normal patients had HSF and TOP performed in random order. Quantitative comparisons for the global visual field indices - mean deviation and defect (MD) for HSF and TOP, and pattern standard deviation (PSD) for HSF and loss variance (LV) for TOP - were made using correlation coefficients. Humphrey global parameters were converted to Octopus equivalents, and method comparison analysis was used to determine agreement between the two strategies. Test duration times were compared using t-test. Sensitivity and specificity for these two algorithms were determined according to predetermined criteria. Results: High correlation coefficient values were obtained for MD measurements between HSF and TOP (r=-0.89, P

Uniformity of visual acuity measures in published studies

PURPOSE. To investigate the methods used in contemporary ophthalmic literature to designate visual acuity (VA). METHODS. Papers in all 2005 editions of five ophthalmic journals were considered. Papers were included if (1) VA, vision, or visual function was mentioned in the abstract and (2) if the study involved age-related macular degeneration, cataract, or refractive surgery. If a paper was selected on the basis of its abstract, the full text of the paper was examined for information on the method of refractive correction during VA testing, type of chart used to measure VA, specifics concerning chart features, testing protocols, and data analysis and means of expressing VA in results. RESULTS. One hundred twenty-eight papers were included. The most common type of charts used were described as logMAR-based. Although most (89.8%) of the studies reported on the method of refractive correction during VA testing, only 58.6% gave the chart design, and less than 12% gave any information whatsoever on chart features or measurement procedures used. CONCLUSIONS. The methods used and the approach to analysis were rarely described in sufficient detail to allow others to replicate the study being reported. Sufficient detail should be given on VA measurement to enable others to duplicate the research. The authors suggest that charts adhering to Bailey-Lovie design principles always be used to measure vision in prospective studies and their use encouraged in clinical settings. The distinction between the terms logMAR, an acuity notation, and Bailey-Lovie or ETDRS as chart types should be adhered to more strictly. Copyright © Association for Research in Vision and Ophthalmology.

Change of visual acuity recording methods in clinical studies across the years

Purpose: To report any differences in the visual acuity (VA) recording method used in peer-reviewed ophthalmology clinical studies over the past decade. Methods: We reviewed the method of assessing and reporting VA in 160 clinical studies from 2 UK and 2 US peer-reviewed journals, published in 1994 and 2004. Results: The method used to assess VA was specified in 62.5% of UK-published and 60% of US-published papers. In the results sections of the UK publications the VA measurements presented were Snellen acuity (n = 58), logMAR acuity (n = 20) and symbol acuity (n = 1). Similarly in the US publications the VA was recorded in the results section using Snellen acuity (n = 60) and logMAR acuity (n = 14). Overall 10% of the authors appeared to convert Snellen acuity measurements to logMAR format. Five studies (3%) chose to express Snellen-type acuities in decimal form, a method which can easily lead to confusion given the increased use of logMAR scoring systems. Conclusion: The authors recommend that to ensure comparable visual results between studies and different study populations it would be useful if clinical scientists worked to standardized VA testing protocols and reported results in a manner consistent with the way in which they are measured. Copyright © 2008 S. Karger AG.

Adaptive NormalHedge for robust visual tracking

In this paper, we propose a novel visual tracking framework, based on a decision-theoretic online learning algorithm namely NormalHedge. To make NormalHedge more robust against noise, we propose an adaptive NormalHedge algorithm, which exploits the historic information of each expert to perform more accurate prediction than the standard NormalHedge. Technically, we use a set of weighted experts to predict the state of the target to be tracked over time. The weight of each expert is online learned by pushing the cumulative regret of the learner towards that of the expert. Our simulation experiments demonstrate the effectiveness of the proposed adaptive NormalHedge, compared to the standard NormalHedge method. Furthermore, the experimental results of several challenging video sequences show that the proposed tracking method outperforms several state-of-the-art methods.

Robust visual tracking using structurally random projection and weighted least squares

Sparse representation based visual tracking approaches have attracted increasing interests in the community in recent years. The main idea is to linearly represent each target candidate using a set of target and trivial templates while imposing a sparsity constraint onto the representation coefficients. After we obtain the coefficients using L1-norm minimization methods, the candidate with the lowest error, when it is reconstructed using only the target templates and the associated coefficients, is considered as the tracking result. In spite of promising system performance widely reported, it is unclear if the performance of these trackers can be maximised. In addition, computational complexity caused by the dimensionality of the feature space limits these algorithms in real-time applications. In this paper, we propose a real-time visual tracking method based on structurally random projection and weighted least squares techniques. In particular, to enhance the discriminative capability of the tracker, we introduce background templates to the linear representation framework. To handle appearance variations over time, we relax the sparsity constraint using a weighed least squares (WLS) method to obtain the representation coefficients. To further reduce the computational complexity, structurally random projection is used to reduce the dimensionality of the feature space while preserving the pairwise distances between the data points in the feature space. Experimental results show that the proposed approach outperforms several state-of-the-art tracking methods.

A role for local mechanisms in perceived duration of brief visual events

Background: Spatially localized duration compression of a briefly presented moving stimulus following adaptation in the same location is taken as evidence for modality-specific neural timing mechanisms.

Aims: The present study used random dot motion stimuli to investigate where these mechanisms may be located.

Method: Experiment 1 measured duration compression of the test stimulus as a function of adaptor speed and revealed that duration compression is speed tuned. These data were then used to make predictions of duration compression responses for various models which were tested in experiment 2. Here a mixed-speed adaptor stimulus was used with duration compression being measured as a function of the adaptor’s ‘speed notch’ (the removal of a central band from the speed range).

Results: The results were consistent with a local-mean model.

Conclusions: Local-motion mechanisms are involved in duration perception of brief events.

Methodological variations in estimating apparent progressive visual field loss in clinical trials of glaucoma treatment.

OBJECTIVES:

To compare methods to estimate the incidence of visual field progression used by 3 large randomized trials of glaucoma treatment by applying these methods to a common data set of annually obtained visual field measurements of patients with glaucoma followed up for an average of 6 years.

METHODS:

The methods used by the Advanced Glaucoma Intervention Study (AGIS), the Collaborative Initial Glaucoma Treatment Study (CIGTS), and the Early Manifest Glaucoma Treatment study (EMGT) were applied to 67 eyes of 56 patients with glaucoma enrolled in a 10-year natural history study of glaucoma using Program 30-2 of the Humphrey Field Analyzer (Humphrey Instruments, San Leandro, Calif). The incidence of apparent visual field progression was estimated for each method. Extent of agreement between the methods was calculated, and time to apparent progression was compared.

RESULTS:

The proportion of patients progressing was 11%, 22%, and 23% with AGIS, CIGTS, and EMGT methods, respectively. Clinical assessment identified 23% of patients who progressed, but only half of these were also identified by CIGTS or EMGT methods. The CIGTS and the EMGT had comparable incidence rates, but only half of those identified by 1 method were also identified by the other.

CONCLUSIONS:

The EMGT and CIGTS methods produced rates of apparent progression that were twice those of the AGIS method. Although EMGT, CIGTS, and clinical assessment rates were comparable, they did not identify the same patients as having had field progression.

Promoção da acessibilidade em televisão interativa (iTV) : proposta de um serviço adaptado a utilizadores com défice visual

As infraestruturas de televisão interativa atualmente existentes possibilitam a integração de uma grande variedade de recursos e serviços, possibilitando aos utilizadores novas experiências de interação e participação. Para a maioria dos telespetadores, o uso de serviços interativos não acarreta grandes dificuldades; no entanto, para públicos com necessidades especiais, por exemplo para pessoas com défice visual, essa tarefa torna-se complexa, dificultando, ou mesmo impedindo, que estes utilizadores possam beneficiar deste tipo de serviços. Portugal não é uma exceção neste contexto, existindo um número significativo de utilizadores com défice visual (UDV) que não beneficiam totalmente das potencialidades do paradigma televisivo atual. Neste âmbito, o projeto de investigação que suporta esta tese explora a problemática do Design Universal aplicado à Televisão Interativa (iTV) e tem como objetivos a conceptualização, prototipagem e validação de um serviço de iTV adaptado especificamente a UDV, visando promover a sua inclusão digital. Para cumprir estes objetivos, a investigação dividiu-se em três etapas distintas. Na primeira etapa, a partir da Teoria Fundamentada nos Dados, foram identificadas as dificuldades e necessidades dos UDV enquanto consumidores de conteúdos televisivos e serviços de audiodescrição; foi selecionada a plataforma tecnológica mais adequada para o suporte do serviço prototipado; e foi definido um conjunto de princípios orientadores de design (POD’s) de interfaces de televisão interativa específico para este público-alvo. Inicialmente foram efetuadas duas entrevistas a 20 participantes com défice visual, para determinar as suas dificuldades e necessidades enquanto consumidores de conteúdos televisivos e serviços de audiodescrição. De seguida, foi realizada uma entrevista a um perito responsável pelo processo de transição para a TDT em Portugal (inicialmente considerou-se que a TDT seria uma plataforma promissora e poderia suportar o protótipo) e efetuada a revisão da literatura sobre POD’s para o desenvolvimento de interfaces para serviços iTV dirigidos a pessoas com défice visual. A partir dos resultados obtidos nesta etapa foi possível definir os requisitos funcionais e técnicos do sistema, bem como os seus PODs, tanto ao nível da componente gráfica, como de interação. Na segunda etapa foi concetualizado e desenvolvido o protótipo iTV adaptado a UDV ‘meo ad+’, com recurso à plataforma tecnológica IPTV da Portugal Telecom, seguindo os requisitos e os princípios de design definidos. Relativamente à terceira etapa, esta contemplou a avaliação do serviço prototipado, por parte de um grupo de participantes com défice visual. Esta fase do trabalho foi conduzida através do método de Estudo Avaliativo, possibilitando, através de testes de usabilidade e acessibilidade, complementados com entrevistas, compreender se o serviço prototipado ia efetivamente ao encontro das necessidades deste tipo de utilizadores, tendo-se observado que os participantes que estiveram envolvidos nos testes ao protótipo mostraram-se satisfeitos com as funcionalidades oferecidas pelo sistema, bem como com o design da sua interface.

Visual cortex frontend: integrating lines, edges, keypoints and disparaty

We present a 3D representation that is based on the pro- cessing in the visual cortex by simple, complex and end-stopped cells. We improved multiscale methods for line/edge and keypoint detection, including a method for obtaining vertex structure (i.e. T, L, K etc). We also describe a new disparity model. The latter allows to attribute depth to detected lines, edges and keypoints, i.e., the integration results in a 3D \wire-frame" representation suitable for object recognition.