With advances in space-borne imaging sensor technology, the potential for building geospatial databases has increased immensely; yet extracting target information from the high resolution remotely sensed images still poses a challenging problem. Many satellites have been launched to analyse different terrain regions such as deserts, coastal areas, metro's, vegetation, to carry out geological surveys for environmental monitoring, disaster forecasting etc. These satellites acquire a large number of images everyday leading to an exponential increase in the database containing unstructured and unorganized satellite images. Hence there is imminent need to develop a robust system to retrieve a set of images from this unstructured database that will meet the user's requirement. Conventional query processing systems are based on matching keywords such as time of acquisition, geographic locations, sensor types, etc. The standard dataset which is available on www.earthexplore.nasa.us.gov.in/images containing 1000 images is used to carry out the research. A detailed comparative analysis of retrieved results in accuracy, precision, recall and F-measure is carried out in each phase of the work.