Visual Information and Information Systems

Frontmatter

Unsupervised Color Film Restoration Using Adaptive Color Equalization

Abstract

Chemical processing of celluloid based cinematic film, becomes unstable with time, unless they are stored at low temperatures. Some defects, such as bleaching on color movies, are difficult to solve using photochemical restoration methods. In these cases, a digital restoration tool can be a very convenient solution. Unfortunately, for old movies color and dynamic range digital restoration is usually dependent on the skill of trained technicians who are able to control the parameters through color adjustment, and may be different for a sequence or group of frames. This leads to a long and frustrating restoration process. As an alternative solution, we present in this paper, an innovative technique based on a model of human color perception:, to correct color and dynamic range with no need of user supervision and with a very limited number of parameters. The method is combined with a technique that is able to split the movie into different shots and to select representative frames (key frames) from each shot. By default, key frames are used to set the color correction method parameters that are then applied to the whole shot. Due to the robustness of the color correction method the setting used for the key frame is used successfully for all the frames of the same shot.

A. Rizzi, C. Gatta, C. Slanzi, G. Ciocca, R. Schettini

Grey-Scale Image Colorization by Local Correlation Based Optimization Algorithm

Abstract

In this paper, we present a grey-scale image colorization technique by using local correlation based optimization algorithm. The core of our colorization method is to formalize the colorization problem as minimizing a quadratic cost function under some assumptions that are mainly based on the local image characters. It can be successfully applied in colorization a variety of grey-scale images. In our colorization method, users only need to freely scribble the desired color in the input grey-scale image, which is a great improvement upon the traditional manual colorization techniques. By introducing new local connectivity factor and distance factor, our approach can effectively alleviate the color diffuseness in different regions, which is one of main problem in previous colorization methods. Additionally, by exploiting subsampling in YUV space, we accelerate the colorization process with nearly the same good results. Experiments show that better colorization results can be obtained faster with our method.

Dongdong Nie, Lizhuang Ma, Shuangjiu Xiao, XueZhong Xiao

Face Recognition Using Modular Bilinear Discriminant Analysis

Abstract

In this paper, we present a new approach for face recognition, named Modular Bilinear Discriminant Analysis (MBDA). In a first step, a set of experts is created, each one being trained independently on specific face regions using a new supervised technique named Bilinear Discriminant Analysis (BDA). BDA relies on the maximization of a generalized Fisher criterion based on bilinear projections of face image matrices. In a second step, the experts are combined to assign an identity with a confidence measure to each of the query faces. A series of experiments is performed in order to evaluate and compare the effectiveness of MBDA with respect to BDA and to the Modular Eigenspaces method. The experimental results indicate that MBDA is more effective than both BDA and the Modular Eigenspaces approach for face recognition.

Muriel Visani, Christophe Garcia, Jean-Michel Jolion

Computer Vision Architecture for Real-Time Face and Hand Detection and Tracking

Abstract

In this paper we present a computer vision architecture to detect and track the face and hands of a human being in real time from a video sequence captured by a webcam. The architecture has a first preprocessing stage, including a color filtering module, a motion filtering module, a color-based segmentation, a processing channels merge module and, finally, a contour search and discrimination module. The aim of the first stage is to discard the image regions which are highly unlikely to correspond with skin. Thus, the second stage of the architecture is a previously trained Fuzzy ARTMAP multiscale neural network module which only processes those image regions selected by the preprocessing stage, which are fully expected to be skin. The neural networks make the last decision about face and hand detection. After that, the architecture tracks the trajectories which face and hands follow.

D. González-Ortega, F. J. Díaz-Pernas, J. F. Díez-Higuera, M. Martínez-Zarzuela, D. Boto-Giralda

Video Spatio-temporal Signatures Using Polynomial Transforms

Abstract

In this paper we integrate spatial and temporal information, which are extracted separately from a video sequence, for indexing and retrieval purposes. We focus on two filter families that are suitable models of the human visual system for spatial and temporal information encoding. They are special cases of polynomial transforms that perform local decompositions of a signal. Spatial primitives are extracted using Hermite filters, which agree with the Gaussian derivative model of receptive field profiles. Temporal events are characterized by Laguerre filters, which preserve the causality constraint in the temporal domain. Integration of both models gives a spatio-temporal feature extractor based on early vision.. Results encourage our model for video indexing and retrieval.

Carlos Joel Rivero-Moreno, Stéphane Bres

Motion Trajectory Clustering for Video Retrieval Using Spatio-temporal Approximations

Abstract

A new technique is proposed for clustering and similarity retrieval of video motion clips based on spatio-temporal object trajectories. The trajectories are treated as motion time series and modelled using orthogonal basis polynomial approximations. Trajectory clustering is then carried out to discover patterns of similar object motion behaviour. The coefficients of the basis functions are used as input feature vectors to a Self-Organising Map which can learn similarities between object trajectories in an unsupervised manner. Clustering in the basis coefficient space leads to efficiency gains over existing approaches that encode trajectories as point-based flow vectors. Experiments on pedestrian motion data gathered from video surveillance demonstrate the effectiveness of our approach. Applications to motion data mining in video surveillance databases are envisaged.

Shehzad Khalid, Andrew Naftel

Interactive Animation to Visually Explore Time Series of Satellite Imagery

Abstract

Are users able to extract relevant information from an animation? There is mixed evidence about the usefulness of animations in the geosciences, and little is known about how users work with animations. This paper focuses on particular aspects: variables of the temporal dimension of a running animation, the dynamic visualization variables. Research emphasis was on methods to use these variables to encode geodata and to enable the user to manipulate the dynamic properties of the animation by interaction with these variables. Vegetation monitoring has been used as case study. A prototype animation environment, designed to explore large time series, has been qualitatively evaluated by experts in monitoring. Data collection methods used are the think aloud method, interviews and a questionnaire. The evaluation revealed user strategies, tool use and the role of effects of animation use. Results indicate that users are able to extract relevant information in a monitoring context.

Connie A. Blok

An OpenGIS®-Based Approach to Define Continuous Field Data Within a Visual Environment

Abstract

Many continuous phenomena affect everyday human life.In this paper, the issue of continuous field management has been faced under several points of views, from the design to the implementation stage. The resulting environment provides domain experts with a formal visual approach to conceptually depict mini-worlds of interest embedding spatial data. It is based on an extended version of the OpenGIS standard architecture, thus guaranteeing a natural integration with solutions derived by previous conventional approaches.

Luca Paolino, Monica Sebillo, Genoveffa Tortora, Giuliana Vitiello

Wayfinding Choreme Maps

Abstract

This contribution details how conceptual characterizations of route knowledge can provide the basis for graphical route information in a cognitively adequate way. The approach is based on the theory of wayfinding choremes that originated from the leitmotif to reflect mental conceptualization processes—as a canonical representation—in different modes of externalization, primarily graphical and verbal. The approach is therefore termed cognitive conceptual approach to map design; it stands in opposition to more frequently used data driven approaches. Possibilities and requirements of the conceptual approach are explored and related to information system requirements such as the semantic specification of data structures and their relation to visual output. The wayfinding choreme approach has been implemented in a basic version; its requirements are illustrated and future lines of research are discussed. The focus is placed on organizational aspects of route knowledge, i.e. how they can be modeled and how they can be accounted for in the visualization of modern navigation assistance systems.

Alexander Klippel, Kai-Florian Richter, Stefan Hansen

An Approach to Perceptual Shape Matching

Abstract

In contrast to geometric similarity, perceptual similarity is manifold and much more difficult to handle. Relatively few work is known towards capturing the perceptual similarity. In this work we introduce the concept of local shape width and show that it can be applied to measure the perceptual significance of shape parts. Given such a measure, one can define the influence of the parts as a function of their significance. By doing so, we tend to base the shape matching on perceptually more meaningful parts and reduce the matching relevance of other parts. As an application, we propose a shape evolution approach, which synthesizes a series of new shapes from an input shape. They have the property that perceptually less significant parts smoothly vanish while other parts remain unchanged. This tool can be used in combination with any shape matching algorithm. A second application is proposed to perform non-uniform shape sampling. Experimental results will be given to show the practical usefulness of the local shape width concept.

Xiaoyi Jiang, Sergej Lewin

Interactive Volume Visualization Techniques for Subsurface Data

Abstract

In this paper we describe concepts, which support the interactive exploration of subsurface information extracted from seismic datasets. Since in general subsurface information is of volumetric nature, appropriate visualization techniques are needed to provide an insight view to special regions of interest. Usually clipping planes or surface extraction techniques are used for this purpose. We will present an approach, which allows the user to interactively change the visual representation for distinct regions of seismic datasets. Using this technique highlighting of regions of interest as well as clipping against volumetric regions can be realized. Volumetric clipping regions have the potential to assist the user when visually intruding into a 3D dataset by permitting an occlusion free view to inner regions of the dataset. During this process it is desirable to know where the current position is located relative to the whole dataset. We will introduce a 3D widget, which displays information concerning the location and orientation of the virtual camera during the exploration process.

Timo Ropinski, Klaus Hinrichs

Compressed Domain Image Retrieval Using JPEG2000 and Gaussian Mixture Models

Abstract

We describe and compare three probabilistic ways to perform Content Based Image Retrieval (CBIR) in compressed domain using images in JPEG2000 format. Our main focus are arbitrary non-uniformly textured color images, as can be found, e.g., in home user image collections. JPEG2000 offers data that can be easily transferred into features for image retrieval. Thus, when converting images to JPEG2000, feature extraction comes at a low cost. For feature creation, wavelet subband data is used. Color and texture features are modelled independently and can be weighted by the user in the retrieval process. For texture features in common databases, we show in which cases modelling wavelet coefficient distributions with Gaussian Mixture Models (GMM) is superior in to approaches with Generalized Gaussian Densities (GGD). Empirical tests with data collected by non-expert users evaluate the usefulness of the ideas presented.

Alexandra Teynor, Wolfgang Müller, Wolfgang Kowarschick

Indexing and Retrieving Oil Paintings Using Style Information

Abstract

In this paper we discuss the principle of color perception in oil paintings, and analyze the existing color spaces used in digital image processing. Because of the complexity of color perception, there is no single color space which can represent various properties of color perception and provide a consistent retrieval capability in indexing oil paintings. We propose a new color space for indexing and retrieving oil paintings, which is based on pigment color mixing and the psychology of seeing. The new space includes seven color elements: pure color, white, black, tint, tone, shade and grey. Finally, six semantic query categories are defined in accordance with six color arrangements in oil paintings.

Yan Yan, Jesse S. Jin

Semi-automatic Feature-Adaptive Relevance Feedback (SA-FR-RF) for Content-Based Image Retrieval

Abstract

The paper proposes an adaptive retrieval approach based on the concept of relevance-feedback, which establishes a link between high-level concepts and low-level features. The user’s feedback is used not only to assign proper weights to the features, but also to dynamically select them and to identify the set of relevant features according to a user query, maintaining at the same time a small sized feature vector to attain better matching and lower complexity. Results achieved on a large image dataset show that the proposed algorithm outperforms previously proposed methods. Further, it is experimentally demonstrated that it approaches the results obtained by optimum feature selection techniques having complete knowledge of the data set.

Anelia Grigorova, Francesco G. B. De Natale

A Visual Query Language for Uncertain Spatial and Temporal Data

Abstract

Query languages for sensor data will have similarities with traditional query languages but will also have diverging properties that cause a higher complexity than the traditional ones. Both types require data independence. However, as different sensors create data of heterogeneous type the commonly used methods for data selection cannot be used. Furthermore, sensor data will always be associated with uncertainties and since also sensor data fusion must be possible to carry out this cause further problem in development of the query languages. Here a visual query language for sensor data input is discussed from these perspectives to allow a complete set of spatial temporal queries by means of its visual user interface.

Karin Silvervarg, Erland Jungert

Surveying the Reality of Semantic Image Retrieval

Abstract

An ongoing project is described which seeks to add to our understanding about the real challenge of semantic image retrieval. Consideration is given to the plurality of types of still image, a taxonomy for which is presented as a framework within which to show examples of real ‘semantic’ requests and the textual metadata by which such requests might be addressed. The specificity of subject indexing and underpinning domain knowledge which is necessary in order to assist in the realization of semantic content is noted. The potential for that semantic content to be represented and recovered using CBIR techniques is discussed.

Peter G. B. Enser, Christine J. Sandom, Paul H. Lewis

Too Much or Too Little: Visual Considerations of Public Engagement Tools in Environment Impact Assessments

Abstract

Recently proposed reclamation works due to take place in the Victoria Harbor of Hong Kong have raised questions about their appropriateness and desirability. Although the plans for reclamation had gone through the Environmental Impact Assessment (EIA) process and the submitted report available online, its wordy and technical contents were not well received by the public. The report failed to offer the community at large a better understanding of the issues at hand and to visualize what would become of the proposed site upon project completion. Henceforth, the Environmental Protection Department stipulates that future EIA reports be presented in a format more readily comprehensible than written accounts. This requirement calls for more visual displays, including but not limited to, three dimensional models, maps and photo imageries. In compliance with the requirements and recognizing technological impetus, we structured a web-based platform that makes use of the Geographic Information System technology to explore alternative visual presentation, such as maps, graphics, photos, videos, and animations. Our research has demonstrated that visual resources are viable substitutes to written statements in conveying environmental problems albeit with limitations. This paper shares our knowledge and experience in compiling visual resources and hopes that our integrative effort is a step forward in the development of a more effective public engagement tool.

Ann Shuk-Han Mak, Poh-Chin Lai, Richard Kim-Hung Kwong, Sharon Tsui-Shan Leung

Active Landmarks in Indoor Environments

Abstract

Landmarks are an important enhancement for pedestrian navigation systems. They are not only aids at decision points but they are also an affirmation to the user that he is still on the correct route.

Especially in indoor environments where the density of conventional landmarks is rather low, the implementation of so called “Active Landmarks” is an enrichment to the system. This notation derives from the fact that information is actively sent to the handheld device without any user interaction. That way the user receives information from and about a specific landmark, especially concerning its position and consequently the user’s position. This is particularly important within buildings where the user needs detailed position information. Because of the lack of outstanding elements, there are many possibilities to get lost and orientation is much more difficult than outdoors. Nevertheless positioning techniques are scarcely offered and in case where they are available their usage is cost intensive or not accurate enough. In this article the importance of Active Landmarks and their implementation with the help of new technologies like RFID is discussed.

Beatrix Brunner-Friedrich, Verena Radoczky

Image Annotation for Adaptive Enhancement of Uncalibrated Color Images

Abstract

The paper describes an innovative image annotation tool, based on a multi-class Support Vector Machine, for classifying image pixels in one of seven classes – sky, skin, vegetation, snow, water, ground, and man-made structures – or as unknown. These visual categories mirror high-level human perception, permitting the design of intuitive and effective color and contrast enhancement strategies. As a pre-processing step, a smart color balancing algorithm is applied, making the overall procedure suitable for uncalibrated images, such as images acquired by unknown systems under unknown lighting conditions.

Claudio Cusano, Francesca Gasparini, Raimondo Schettini

Automatic Redeye Removal for Smart Enhancement of Photos of Unknown Origin

Abstract

The paper describes a modular procedure for automatic correction of redeye artifact in images of unknown origin, maintaining the natural appearance of the eye. First, a smart color balancing procedure is applied. This phase not only facilitates the subsequent steps of processing, but also improves the overall appearance of the output image. Combining the results of a color-based face detector and of a face detector based on a multi-resolution neural network the most likely facial regions are identified. Redeye is searched for only within these regions, seeking areas with high “redness” satisfying some geometric constraints. A novel redeye removal algorithm is then applied automatically to the red eyes identified, and opportunely smoothed to avoid unnatural transitions between the corrected and original parts. Experimental results on a set of over 450 images are reported.

Francesca Gasparini, Raimondo Schettini

Analysis of Multiresolution Representations for Compression and Local Description of Images

Abstract

Low level features of images are often extracted from their representations in a Gaussian scale space. These representations satisfy the desired properties of covariance under a set of transformations (translations, rotations, scale changes) as well as of causality. However, the corresponding image representations, due to their redundant and non sparse nature, are not well suited for compression purposes. This paper aims at characterizing a set of multiresolution representations from the joint perspective of feature point and descriptor extraction and of compression. This analysis leads to the design of a feature point detector and of a local descriptor in signal representations given by oversampled steerable transforms. It is shown that the steerable transforms due to their properties of covariance under translations, and rotations as well as of angular selectivity provide signal representations well suited to address the signal description problem. At the same time, techniques such as iterative projection algorithms (POCS – projection on convex sets) are used to reduce the coding cost induced by the corresponding oversampled signal representation. The robustness and the discriminative power of extracted features are rated in terms of the entropy of the quantized representation. These results show the tradeoff that can be found between compression and description.

François Tonnin, Patrick Gros, Christine Guillemot

Presenting a Large Urban Area in a Virtual Maquette: An Integrated 3D Model with a ‘Tangible User Interface’

Abstract

Providing information and combining information to make a decision on the most suitable location for locating a new building or office is not a new concept. However the approach which was used in the Holland Promotion Circle (HPC) is entirely new. Instead of paper maps and analogue 3D models, a tangible interface, augmented reality and a 3D virtual environment was used to provide information for foreign investors that are considering locating their business in the Randstad area of the Netherlands. To make a selection of suitable data on the database, profile recognition through RFID was used. The Pilot period of the completion of the HPC opened the possibilities to find out the advantages of using a TUI and to experiment with the use of an enormous dataset. How to cope with the fast amounts of data and how to present the appropriate data to the visitors of the HPC.

Irene Pleizier, Evert Meijer

Perceptual Image Retrieval

Abstract

This paper addresses the problem of texture retrieval by using a perceptual approach based on multiple viewpoints. We use a set of features that have a perceptual meaning corresponding to human visual perception. These features are estimated using a set of computational features that can be based on two viewpoints: the original images viewpoint and the autocovariance function viewpoint. The set of computational measures is applied to content-based image retrieval (CBIR) on a large image data set, the well-known Brodatz database, and is shown to give better results compared to related approaches. Furthermore, results fusion returned by each of the two viewpoints allows significant improvement in search effectiveness.

Noureddine Abbadeni

Analyzing Shortest and Fastest Paths with GIS and Determining Algorithm Running Time

Abstract

In this paper, a few tests have been performed for determining optimum and faster path in the networks. Determining the shortest or least cost route is one of the essential tasks that most organizations must perform. Necessary software based on CAD has been developed to help transportation planning and rescue examinations in the scope of this research. Two analyses have been performed by using Dijkstra algorithm. The first one is the shortest path which only takes into account the length between any two nodes. The second is the fastest path by introducing certain speeds into the paths between the nodes in certain times. A case study has been carried out for a selected region in Istanbul to check the performance of the software. After checking the performance of the software, running time of used algorithm was examined. According to the established networks, which have different nodes, the behaviors of algorithm running time were determined separately. Finally, the most appropriate curve was fitted by using CurveExpert 1.3 program according to algorithm running times in different networks.

Turan Erden, Mehmet Zeki Coskun

Multimodal Data Fusion for Video Scene Segmentation

Abstract

Automatic video segmentation into semantic units is important to organize an effective content based access to long video. The basic building blocks of professional video are shots. However the semantic meaning they provide is of a too low level. In this paper we focus on the problem of video segmentation into more meaningful high-level narrative units called scenes – aggregates of shots that are temporally continuous, share the same physical settings or represent continuous ongoing action. A statistical video scene segmentation framework is proposed which is capable to combine multiple mid-level features in a symmetrical and scalable manner. Two kinds of such features extracted in visual and audio domain are suggested. The results of experimental evaluations carried out on ground truth video are reported. They show that our algorithm effectively fuses multiple modalities with higher performance as compared with an alternative conventional fusion technique.

Vyacheslav Parshin, Aliaksandr Paradzinets, Liming Chen

Backmatter