Development of descriptors for natural feature identification on IRS LISS III images

CBIR is a technique which aims at retrieving images which are perceptually and semantically meaningful by making the best use of a single or combination of visual content descriptors. The application domain of this technique generally excludes satellite images, but this paper explores the usability of CBIR on satellite imagery obtained from well known Indian Remote Sensing (IRS LISS III sensor) satellite. One can ponder over the contribution of remotely sensed data towards diverse scientific and commercial purposes. Hence, it is reasonable to selectively identify and characterize the natural surface features, e.g., hills, rivers, trees, fields, forests and soils and recognise objects like aeroplane, oil tank, ships, etc. The interpretation and analysis of such features is a major step in transforming the remotely sensed data into information that is intelligible and usable. Object/region identification and search form challenging tasks in image interpretation which is itself a hot research area under the broad spectrum of CBIR. Computer-assisted search for an object can be done in an image by extracting one or more salient characteristics like shape, color, texture, and pattern. This paper presents a pixel by pixel search of exactly or relatively similar occurrences of the queried region/object considering a fusion of two prominent descriptors namely, texture and color as search criteria. The task has been done in a distributed fashion.