Research On Automatic Image Mosaic Technology Of Large Format CIS Scanner

Due to the constraints of production process and cost, the length of a single CIS(Contact Image Sensor) is difficult to achieve the requirements of A0 and the above format. Therefore, a majority of large format CIS scanners are usually spliced by several A4 width CIS. So the high precision image mosaic technology is becoming a key technology to obtain high quality images. In the past and present, semiautomatic mosaic technology which needs interaction of manual and computer was/is widely used. In recent years, with the rapid development of information and automation technology, higher requirements have been put forward to the automatization and intelligentization of next generation scanner.The main purpose of this thesis is to study the real-time automatic image mosaic technology of a new generation large format CIS scanner, not only to achieve the automatic calculation of mosaic parameters, but also to achieve real-time image mosaic. Based on this purpose, a novel method based on image registration algorithm and FPGA dynamic image mosaic is presented by this thesis. The first step is to scan a drawing with rich feature, automatically calculate the mosaic parameters through the software algorithm for feature extraction and feature matching. The next step, the computer will send the mosaic parameters to the scanner, and the scanner will store the data. Finally, during the normal scan process, the scanner will splice the scanning image directly according to the stored parameters. This method makes full use of the high precison characteristic of image registration and the real-time characteristic of FPGA. In this way, automatic calculation of mosaic parameters and real-time image mosaic have also been realized. According to our research, no similar method was reported in the refer data about the study of CIS scanner.According to the above research approach, the major work and achievements of this thesis are as follows:(1) The principle of the non-uniformity of the scanner image has been studied, and a multi-section calibration method has been proposed to solve the problem of non-uniformity. The method works well and effectively. Not only the overall quality of the images has been improved, but also the accuracy rate of the feature point detection and matching becomes higher.(2) Three registration methods which are separately based on gray, transform domain and feature have been analysized, and the feature-based image registration method is the main research direction in this thesis. Hence three feature point detection algorithm such as Harris, SIFT and SURF have been studied to compare their performance in different illumination, fuzzy conditon and noise strength. Additonally the matching methods of the three kinds of feature points have also been studied to compare their matching effect. Finally the SURF algorithm which is proved very superior has been choosen to complete the image registration. The relative position of the CIS and mosaic parameters is calculated according to the image registration results.(3) A detailed image mosaic scheme which is based on FPGA and SDRAM has been made. Firstly, FPGA receives and keeps the mosaic parameters calculated by software. Sencondly, the storage stucture and read-write timing of SDRAM will be controlled by FPGA according to the mosaic parameters. Thus image mosaic can be completed in the scanner hardware.A method which is composed of the image registration algorithm based on feature points and dynamic image mosaic based on FPGA is proposed in this thesis. This method has already been used in the R&D of high precision large format CIS scanner. It has prompted the realization of automatic calculation of mosaic parameters and real-time image mosaic. In the refer data about the image mosic technology of large format CIS scanner, this method is firstly reported in China.