Algorithms and Results in Content-Based Visual Query of the Image Databases Resulting from Dicom Files

Abstract

The article presents a series of algorithms used in the content-based visual query process on databases with color medical images extracted from DICOM files provided by medical tools used in the diagnosis area. Also, the paper presents a detailed study on several directions inside the content-based visual query process. In studying the content-based image query on color feature the transformation from RGB color space to HSV and the quantization at 166 colors was used. For computing the dissimilitude between the query and the target image three metrics were implemented: the Euclidian distance, the histograms intersection and the quadratic distance between histograms. The next study was made on content-based image query on color texture feature. In order to compute the color texture characteristics vectors two methods were implemented: the co-occurrence matrices and Gabor filters. The third studied problem was the content-based region query. For detecting color regions the color set back-projection algorithm was implemented. Statistic studies for this algorithm in keeping track of the patient evolution during the treatment of the peptic ulcers were made