Creating Image Mosaics using Statistical Methods for Images Alignment

Abstract<br /> Image mosaics is a collection of small images arranged in a way that when they are seen from a distance they seen as larger image. It occupies a wide field in computer vision and image processing area. This field is performed in four stages. These stages are selecting the target image (input image) stage, choosing the tiling pattern stage. These tiles are small images from image database. The third stage arranged the tiles in a way to cover the target image. Indeed, color correction stage where the image colors are adjusted to better match the target image colors. The aim of our work is to create an image mosaics based on Statistical approaches.<br /> Database images are collected and selected randomly. Then these images are pr-processed through a resized operation. Also a color transformation from RGB to the grayscale form (intensity). The color transformation is necessary because the intensity form shows the details of the target and the tile images to obtain effective images matching. At first we select a target image randomly. The target image is segmented to 32x32 tiles. At the second stage we choose the rectangular pattern for its simplicity as well as the rectangular nature of tile images used. Arranging tile stage is obtained, for each tile the system search the image database and selects the best matching image based on correlation coefficients of the intensity images. At this stage the difference between our works than others appeared. We use the correlation coefficients for matching operation that obtain a best matching than other approaches like Mathematical approaches which obtained approximate results such as distance functions. Also other Statistical approaches such as Histogram intersection and Statistics measure (Chi Square) used to arrange tiles operation. Finally, we perform image color correction in pixel wise using the average of the R, G, B channels of the target and tile images. The resultant pixel obtained from last step will cover the target. As a result an image mosaic is created. The proposed work shows efficient and accurate results.