Feature extraction of underwater images using principle component analysis with image registration

Abstract Digital imaging becomes the foremost important techniques in environmental monitoring and exploration. Underwater images are used to assess variety and profusion of objects are adopted by marine biologists. The quality of underwater image is necessary in the field of deep-sea study for object annotations, identifications and classifications. Automatic processing of analyzing the images working to attain the objectives in reduced price and period. There are many challenges involved such as complex background, deformation, low resolution and light propagation. With the advancement automatic detection, the deep neural network has led in real-time scenarios to detect and recognize the objects. We used the basic image processing operations such as registration and registered and fusion before performing the feature extraction. By using these techniques the proposed system takes advantages and extracted more features. The Harries Detection method is used for detecting the matched points and principal component analysis is used for feature matching. We have implemented with MATLAB and compared results with the existing systems by calculating the Entropy (E), Root Mean Square Error (RMSE) and Peak Signal Noise Ratio (PSNR). The proposed system shows 98% of accuracy compared to the existing systems.