Research On Optimal Stitching Line And Color Blending Algorithm In Image Stitching

Image stitching is a hot topic in the field of computer vision.Its main purpose is to solve the problem that happens because of the limitations of photographic equipment,that is,a larger view of the image can't be directly obtained.As an important branch of image processing,the advancement of image stitching is helpful for the development of intelligent cars,medial image synthesis,and military remote sensing image synthesis,etc.When shooting the same scene from different perspectives,there are often exposure differences and object parallax amount multiple images.Using the traditional stitching methods will make images appear in unnatural color transitions and blurred in the overlapped area.Moreover,if there are moving objects in the scene,it is more likely to cause ghosts in mosaic images.In order to eliminate ghosts,we propose a global optimized search algorithm based on the traditional dynamic programming method.In addition,for the problem of brightness inconsistency caused by exposure differences,we propose a color interpolation model based on optimal stitching line.Their specific ideas are briefly introduced as follows:1.In order to obtain a stitching line that can successfully avoid moving objects,we first construct a new energy function based on color and texture similarity.In this paper,for the current point,we search for all points rather than only three neighboring points in its next row.Then,we find a global optimal stitching line on our energy function by using the improved algorithm.By stitching the aligned images together along this stitching line,it is possible to avoid moving objects and achieve a no-ghost image.2.In order to deal with the problem of brightness inconsistency in the image,we propose a fast color interpolation model based on the MVC and SLIC.We calculate a color difference value for all points on the right side of the stitching line,and then add them corresponding to the original pixel to eliminate the brightness inconsistency.In addition,in the process of experiment,we speed up our fusion efficiency by using segmentation technique to divide the target image into a number of superpixels,and then using the same color difference value for all pixels in each superpixel to execute color interpolation.In summary,various experimental results show that our stitching results are more natural in visual sense.In addition,the data in this paper have verified the validity of the proposed methods.