Image Stitching Based On The Mirco-camera Array Imaging System

According to the traditional optical theory,there is a restriction between the large field of view and the high resolution.Using the imaging system based on the micro-camera array is one of the effective methods to solve this problem.Through the photos collected from different image sensor,a specific algorithm is designed to generate a picture of high resolution and wild field of view by the overlapping area of these photos.The image stitching algorithm affects the final image quality very much.There are three main steps of the image stitching: image preprocessing,image registration and image fusion.In image preprocessing,geometric correction is used to remove the geometric distortion caused by different view of the sensor,and the noise in the image is smoothed by the filtering algorithm.The accuracy of image stitching will be improved after image preprocessing.Image registration is the process of finding the transformation relationship of the image to be located in right space,by looking for feature space,similarity measure,searching space and searching strategy.The purpose of image fusion is to eliminate the splicing of the stitching image in the result,to make it an image.As the main part of the image stitching,the selection of the image registration algorithm determines the quality of the final stitching result to a great extent.Because of its rotation and scale invariance and good robustness,SIFT is the most common image registration algorithm.But the slower calculation speed and the larger consumption of storage make SIFT hard to be used in some large scale and real time image stitching system.In order to solve this problem,some image registration algorithms based on binary descriptors have been proposed in recent years.These algorithms use the binary string to describe the feature points,compared to the traditional algorithm of high dimensional vector,significantly reducing the storage space and improving the computing speed.For those image systems which requires speed much than accuracy,the image stitching algorithm based on the binary descriptors has a good performance.Due to masses of calculation,an adaptive image stitching algorithm based on SURF and BRISK is designed to accelerate the computation and reduce the need of space.According to the design of optical system,the image system based on micro-camera array adopts CMV4000 image sensor with the resolution of 2048 × 2048.A hardware drive circuit is designed with XQR2V3000 FPGA.The signal collected from the image sensor is transmitted to the back-end signal processing circuit after serial and parallel conversion.The algorithm has good effect through imaging experiment and has some guiding effect on practical engineering.