Advances in Water Resource and Protection (AWRP) Volume 1 Issue 2, April 2013 www.seipub.org/awrp 21 Identification of Water Bodies from High and Multi-level Resolution Satellite Images Using Novel Feature Extracting Technique Satellite Image Processing S Raja Mohamed 1 , YSSR Murthy 2 1 KIT, Coimbatore, TN, India; 2 SVECW, Bhimavaram, AP, India 1 raja_delip@yahoo.com; 2 yssrmoorthy@gmail.com Abstract Water bodies are precious resources which play an important role in agriculture, transportation, drinking water supply and environmental balance of ground water level and so on. Assessment on water-bodies as well as escribing on its characteristics is the most crucial work to be done in India where land is being occupied at a very high rate. Satellite images are not clear all the time and will have shadow effects, which has to identified and distinguished. Water body image may appear similar to shades of huge buildings due to reflection and echoing. Few substantial research have been conducted to extract information about water bodies from several multi-resolution satellite images. This paper proposes a novel methodology based on texture and spectral information of the satellite images to extract water bodies. Keywords Multi-resolution Satellite Image; Remote Sensing Introduction The proposed feature extraction methodology based on textural, spectral and contextual information is described. This methodology, expected to increase the accuracy of extracted single specified feature like water bodies, consists of four modules. Firstly, the best texture element from panchromatic image is extracted using Grey Level Co-occurrence Matrix in overlapping and non-overlapping modes. Secondly, fused textural and multispectral image will be classified using Maximum Likelihood classifier. Thirdly, Edge detector Canny will be applied to provide the contextual data. Finally, textural and contextual data will be given to Markovien Random Field filter to extract the feature. The result of the experiments conducted on IRS P6 LISS IV data is presented. The results of overlapping and non overlapping methods are compared. Background Information It is prominent that no earth observation satellites are orbiting our planet to provide rife imagery of its surface. The space borne sensors can be divided in the measuring reflection of sunlight into the visible and infrared part of the electromagnetic spectrum and thermal infrared radiance, and those actively transmitting microwave pulses and recording the received signal. Optical satellite systems are the most frequently applied in water body extraction research. The parts of the electromagnetic spectrum covered by these sensors include the visible and near infrared (VNIR) ranging from 0.4 to 1.3 μm, the shortwave infrared (SWIR) between 1.3 and 3.0 μm, the thermal infrared (TIR) from 3.0 to 15.0 μm and the long- wavelength infrared (LWIR) from (7-14 μm). Landsat is still among the widest used satellites, partly because it has the longest time series of data of currently available satellites. The first satellites of the Landsat family were equipped with the Multispectral Scanner (MSS), having four bands at 80-m resolution. AVHRR (Advanced Very High Resolution Radiometer) has five bands in 1.1-km resolution and has been flown on many platforms, including TIROS-N (Television Infrared Observation System). Later Landsat satellites had the Thematic Mapper (TM) sensors onboard with improved resolution and more spectral bands. The Indian Remote Sensing Satellites (IRS) 1A and 1B have two sensors called LISS-1 and LISS-2 (Linear Imaging and Self-Scanning Sensor), which are identical except for a two times higher spatial resolution on LISS-2. IRS 1C and 1D also have an identical payload being a 5.8- m resolution panchromatic camera (PAN) and a 23.5- m resolution multispectral sensor called LISS-3.