HOG and SIFT Transformation Algorithms for the Underwater Image Fusion

The information from two or more source images is combined on the basis of information per pixel by characteristic or conclusion level methods in the image fusion technique and the resultant image comes in the form of a single image. The objective of an image fusion algorithm is the integration of the superfluous and matching data gathered from the source descriptions for the formation of a more accurate picture. This image gives an improved explanation of the scene. The underwater image can be enhanced with the help of techniques of white balancing and image blending. The features are matched to carry out the image blending. In this paper, two methods namely, Histogram of Oriented Gradients (HOG) transformation and Scale Invariant Feature Transform (SIFT) are deployed to perform the feature matching. Histogram of oriented gradients (HOG) is a feature descriptor. It is widely used in computer vision and image processing for object detection purposes. SIFT feature representation, coined by Lowe, is a technique of transforming image data into scale-invariant coordinates proportionate to local features. SIFT is based on descriptor histograms of gradients (HOGs), which are calculated in the vicinity of the detected interest points. These feature detectors and descriptors are known to be invariant for a broad spectrum of geometric and photometrical transformations of the images. The HoG and SIFT methods provide haze free images with better color contrast, sharpened features, and preserved edges.