Adaptive Markov Random Field Model for Area Based Image Registration and Change Detection

ABSTRACT Image registration is an important step in image processing task. It is obtained from the combination of two images. Satellite imaging is widely used for various real time applications such as environment monitoring, weather forecasting , medical image processing and in computer vision. These images are captured with the help of different sensors and different acquisition time which affects the performance of real time application. Significant information can be extracted by applying image registration techniques. Various techniques have been proposed by addressing the issue of image registration which is based on feature based image registration and area based image registration. Satellite images have huge amount of data in which feature based image registration techniques fail to perform image registration efficiently. To overcome this, area based image registration technique is proposed. The proposed approach follows Markov random field model for image registration to improve the efficiency of registration. The proposed approach is a modular scheme and does not require any metric computations which help to achieve low complexity during implementation. The model is implemented using panchromatic and multispectral images that acquired from different satellites. The algorithm is experimented for remote sensing, coastal and conventional images. The result shows that proposed method is able to perform efficiently for image registration and is evaluated using the parameters such as Mutual index(MI), Root Mean Square Error(RMSE), Standard Deviation (SD) and Correlation Coefficient(CC). Keywords: Area based registration, Adaptive markov random field, Coastal image, Remote sensing, Satellite system