Object-based image analysis is an approach to remote sensing imagery in which groups of similar pixels, called image-objects, rather than single pixels, are the units analysed.1
Blaschke and co-authors cite a definition of image-objects as basic entities in an image, perceptually generated from high-resolution pixel groups whose data values are similar, and which have an intrinsic size, shape and geographic relationship with the scene component they model.1 The article opens by discussing limitations of per-pixel methods when applied to high resolution images, which is the problem the approach addresses.1
Segmentation is central but not the whole method. The article states that in the geographic form of the approach image segmentation is not an end in itself, and that segmentation is only one, though very typical, delineation strategy.1 It also reports that Hay and Castilla argued in 2008 for including geographic space in the name, giving the abbreviation GEOBIA.1
This page has a single witness among the retained sources, which do not establish the accuracy of object-based classification against pixel-based classification, any particular segmentation algorithm or software, or how segment parameters should be chosen.