Panorama Mosaics Studies Based On Sequence Color Images

Nowadays, images with widefield of view and panoramicim ages are used in more and more applications. They can be produced by some special facilities(such fish-eye camera), whose disadvantages are higher price and more complex manipulation. For this reason, how to get such images with normal cameras has attracted many a tentions. This problem can be resolved image mosaics technique, a branch of image processing, which mosaics acollection of partial overlapped images into a large image with wider field of view.This paper makes the core methods of image mosaics technology—the image registration methods. The overall idea of matching algorithm is to ensure accuracy of matching and small operation. This paper makes some comparative analysis on kinds of image matching methods,that is to say model matching method, adjacent columns ratio matching method, images stitching based on local entropy difference. Experiments prove that these methods always construct template arbitrarily,if this template didn't include enough feature,it will induce failure,so this paper brings forward improved method based on local entropy difference,that is to say before matching,firstly the reference template will be constructed at the position of the feature object area whose location is based on local edge density and fuzzy clustering algorithm,then take it as template to compute it's local entropy difference ,search in the another image to get the best matching point. Through this method images are stitched well.To sequence color images panorama matching,using Microsoft Visual C++6.0 we develop a full process system. Experiments prove that improved algorithm based on local entropy difference can realize panorama matching well.