S. Levachkine, A. Velázquez, and H. Sossa (Mexico), and V. Alexandrov (Russia)
Image Segmentation and Representation, Color and Texture
In this work, we introduce an approach for automatic global thresholding (the threshold is computed in function of the source image) applicable to the case of raster scanned color images. We call this approach the false color technique. We follow the concept that important semantic information necessary to interpret an image is not represented in single pixels but in meaningful image segments and their mutual relations. We exchange the source image by a sequence of images derived from its RGB components. Every image representation thus obtained is associated with either certain color component or a linear function f (R, G, B). We also use transformation of the color image to an intensity image. Some of these representations that contain the objects of interest serve to be used in the following object recognition-interpretation. The method has been tested with complex color images, rendering promising results.
Important Links:
Go Back
