Calibration And Stitching Of Images Scanned By Multiple Image Sensor Arrays

In the field of precision image scanning,the linear sensor,which is represented by the scanner and duplicator,occupies the absolute share.In the aspect of reducing images quality,the linear sensor has great advantage of others.However,in terms of cost and structure,the structure with linear sensor should rely on complex drag mechanism and stringent assembly accuracy.Compared with the former,the size of scanning images is limited by the size of the image sensor array,there are more quality defects in quality,but the image sensor array can do scanning with high speed and low cost without rely on mechanical structure.If the scanning range and the quality of images by image sensor array can be improved,there will be a firm foothold for the image sensor array in image scanning field of miniaturization,high speed and high precision.Based on the experience of the scanner industry,this paper puts forward the structure with multiple image sensor arrays and its image processing flow.New structure not only has the characteristics of the image sensor array itself,but also larger range of scanning,after improving distortion,uneven brightness and stitching quality defects,in the field of small plane scanning and high speed and high precision in will have a place to live in.Scanning structure with multiple image sensor array by adding the number of sensor to extend the field of camera vision.The second chapter describes the structure with four image sensor arrays.through extend the field of camera vision by 4 sensors array.The system improves the image of the defect by means of image enhancement algorithm.At the same time,collecting the images from all the image arrays,and output complete image with high quality.The image processing flow of the new structure divides into two stages: initialization and operation.In the operation phase,the optimized parameters can be calculated within the initial stage,and the time of image enhancement can be great compressed.When the images scanned by multiple image sensor arrays should be stitched,the interlaced image defects will be generated.In the following chapters,the corresponding processing algorithms are proposed for the specific defects of the new structure in the process of image integration.Different from one dimension distortion by the linear sensor,images by the image sensor array will produce two dimensions radial distortion.In order to improve accuracy of the structure,sharpening algorithm is used to extract the corner,and multi view of same images is used to improve the fitting precision of the distortion parameters.The uneven brightness defect of images by the image sensor array will also extend to two dimensions.According to uneven brightness in one image,a method of HSV color space conversion combined with wavelet decomposition is proposed in the fourth chapter,which can reduce uneven brightness defect and maximize maintain chromaticity.The images by the multiple image sensor arrays need to be properly positioned,and overlapped with the overlap region.In the fifth chapter,two kinds of matching algorithms are introduced,which is the feature point affine transformation to the standard coordinates,and the feature points matching between the images.All of them can achieve the matching effect in the model.in the sixth chapter,the defect of edge seam caused by image stitching can be reduced by using of multiple transition functions or wavelet decomposition.This paper will combine with the simulation results before and after the image processing by using of MATLAB algorithm simulation tools in each chapter.The final chapter describes common evaluation system,and clarify existing problems and gives direction for future improvements.