High Definition Image Synthesis Technology Research Of Multi-view Image

With the development of science and the improvement of people's needs for image vision,the high definition image synthesis has become a very important issue.The traditional methods include Inverse Iterative Projection,Dictionary Learning and others which are based on monocular or binocular camera.With the development of image recognition,videos,medical imaging and others,the research of multi-lens has become a hot topic in the field of computer vision.It can make use of the advantages of the multi-lens system to achieve better results and more functions.The research of this paper is the high-definition image synthesis technology of array images,which got by a 3*3 microarray lenses.The main contents of this paper are as follows:(1)It mainly introduces the importance and basic process of high-resolution image.Also it describes some traditional methods of super resolution and image registration and do some comparison experiments.(2)Camera calibration and image registration.In order to get the lenses parameters,we need to calibrate the single lens and stereo calibration.Then,using microarray lens to capture the scene images(9 images).The intrinsic parameters of calibration are used to do the distortion correction,the external parameters are used to replace the sift features as the global variable parameter smooth affine transformation for the registration between 9 images.(3)High definition image synthesis of array images.Using the registered images to do the iterative correction and output an initial high resolution image,and then add the priori information to the initial high resolution image though the method of deep learning,get the final high resolution image.The experimental results show that the high-definition image synthesis method of array images is effective.The advantage of microarray lenses is avoiding the difference time between the traditional multiple camera shooting and it is expected to be applied in handheld electronic device.