Loading...
Please wait, while we are loading the content...
Similar Documents
Extração de características de imagens médicas utilizando wavelets para mineração de imagens e auxílio ao diagnóstico
| Content Provider | Semantic Scholar |
|---|---|
| Author | Silva, Carolina W. Da |
| Copyright Year | 2007 |
| Abstract | Picture Archiving and Communication Systems (PACS) aim at storing all the patients data, including their images, time series and textual description, allowing fast and effective transfer of information among devices and workstations. Therefore, PACS can be a powerful tool on improving the decision making during a diagnosing process. The CAD (Computer-Aided Diagnosis) systems have been recently employed to improve the diagnosis confidence, and recent research shows that they can effectively raise the radiologists performance on detecting anomalies on images. Content-based image retrieval (CBIR) techniques are essential to support CAD systems, and can significantly improve the PACS applicability. CBIR works on raw level features extracted from the images to describe the most meaningful characteristics of the images following a specific criterium. Usually, it is necessary to put together several features to compose a feature vector to describe an image more precisely. Therefore, the dimensionality of the feature vector is frequently large and many features can be correlated to each other. The objective of this Master Dissertation is to build new image features, based on wavelet-generated subspaces. The features form the feature vector, which succinctly represent the images and are used to process similarity queries. The feature vectors are analyzed by the StARMiner system, under development in the GbdI-ICMC-USP, in order to find the most meaningful features to represent the images as well as to find patterns in the images that allow them to be classified into categories. The project developed was evaluated with three different image sets and the results are promising. |
| File Format | PDF HTM / HTML |
| DOI | 10.11606/D.55.2007.tde-07052008-102458 |
| Alternate Webpage(s) | https://teses.usp.br/teses/disponiveis/55/55134/tde-07052008-102458/publico/carol.pdf |
| Alternate Webpage(s) | https://doi.org/10.11606/D.55.2007.tde-07052008-102458 |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
