Exploitable results Image Retrieval Combining Colour and Texture (IMAGEINDEX) The developed image retrieval system allows queries by image example. It considers both colour and texture information. Rotated image cuts can also be used for queries. Due to the increasing amount of digital image data, there is an urgent need for tools to handle that data efficiently. Among other techniques, methods working fully automatically are of special interest for image retrieval from large databases. Colour histograms have proven to be successful in automatic image retrieval, however, a major drawback is that all structural information is lost. Therefore, we extend the colour histogram approach by features that take into account the relations within a local pixel neighbourhood. These features are invariant with respect to translation and rotation. Thus, we combine the advantage of an invariant description (e.g. we only need one histogram for a whole class of transformed images in the database) with the properties of histogram approaches, providing the possibility of also finding images by partial views or vice versa, or to detect objects under occlusion. The disadvantage of histograms, their unsteady assignment at the bin boundaries, is removed by introducing a modified, weighted histogram. Project URL : http://www.informatik.uni-freiburg.de/~noblesse Multi-modality fusion and its application (MMF) Multimedia applications are of prime importance to the technological era. A major point of interest is data fusion. For example in video applications, the two independent channels of image and speech data supply complementary but correlated information. The human observer expects to see and hear corresponding sets of data. When these features are misaligned, say through un-synchronised coding, it is very distracting for the viewer. An understanding of the complex relationship between the two different types of data is essential to ensure successful re-alignment. It will also be of value in very low bit rate coding where it may be possible to transmit only one channel of data and to infer the other from this. A longer term use of information derived from a better understanding of the relationship between image and speech data would be in speech recognition. - The database can be used for studying the relationship between audio and video information. - Use of tracking algorithm for other object extraction. - Use of the multimodality relational information for enhancing communication systems. - Finally, using all the above as add-on units to existing H263, etc. systems. Project URL : http://www.spd.eee.strath.ac.uk/noblesse/ Rational filters for image and video noise smoothing (RF4NS) A family of new nonlinear noise reduction filters, called rational filters, is presented. These filters, which have been applied to both still images and video, present very attractive characteristics in terms of performances and speed versus complexity. Polynomial and rational models have proven to be useful for low-level image processing, in various areas of application (noise smoothing, contrast enhancement, image interpolation). In particular, we address here the problem of edge-preserving noise smoothing in still and moving grey-level or colour images using rational filters. We devised very simple rational operators which outperform other standard linear and nonlinear operators. In fact, their nature allows them to treat data and noise differently, so that they can significantly reduce noise while preserving edges. Moreover, their structure is very simple, so that very economical and fast implementation can be devised, both in the form of ASIC and using Digital Signal Processors. Another advantage of rational filters is that they can cope with different types of noise simply by modifying the values of some parameters, while leaving their structure unchanged. Project URL : http://imagets10.univ.trieste.it/noblesse Automatic and Interactive Segmentation of Video Sequences; AISVS As part of "NOBLESSE", research at the Swiss Federal Institute of Technology has been devoted to the development of non-linear video segmentation schemes. Two different approaches have been investigated. One approach aims at providing automatically a description of the scene as a set of spatio-temporal objects. A spatio-temporal object is a region which presents spatial and temporal homogeneity. It is proven by the results that such a definition provides a semantically meaningful representation of the scene. This is of interest in second generation video coding (e.g. MPEG-4) when the video sequence has to be automatically segmented into semantically meaningful and temporally coherent objects. The second method has been developed in order to provide a flexible tool, capable of addressing a wide range of applications. The system is based on the distinction between regions and semantic objects. The algorithm automatically extracts regions that are homogeneous in colour and/or motion. The grouping of the regions into semantic objects can take place by means of user interaction (e.g. in interactive multimedia applications). However, when the application is known a priori, the grouping can also take place in an automatic mode: for instance, in a surveillance application it can be driven by a change detection mask. Project URL: http://www.ti1.tu-harburg.de/~noblesse Searching for OpenAIRE data... There was an error trying to search data from OpenAIRE No results available
