Algorithm Research On Image Mosaic And Embedded System Realization

Image mosaic is mainly used to synthetize a series of images with the same overlapped scene to an image of large wide viewing angle.After mosaic, image must be close to the original image.Image distortion must be as small as possible and no obvious seams.Image mosaic is widely used in space exploration, undersea exploration, medicine, meteorology, geological survey, military, video compression and transmission, digital preservation of files, video indexing and retrieval, object 3-D reconstruction, military intelligence and public security forensics, digital super-resolution processing, the digital camera.My paper is mainly studying image mosaic algorithm and application in embedded system. Image mosaic background, concept, status and some basic common image mosaic algorithms are discussed in detail.Under the computer simulation in matlab environment,three different image mosaic algorithms from the algorithm accuracy and efficiency factors are compaired. Phase-correlation algorithm has advantages in transplant compared to other algorithm.According to this conclusion, algorithm is transplanted from matlab simulation environment to embedded systems,making it run independently as mosaic image processing applications in the embedded environment. Research focus on:(1)Through comparing computer simulation under matlab environment, template matching method, based on the corner of the image registration method and phase correlation image registration algorithm,the appropriate algorithm ported to embedded systems is found. Although the template matching method can complete the task of image mosaic, it cost much computer times because of finding the phase.Image matching based on feature points can adapt the light conditions, but the calculation is also large.It is hard to do with threshold. It may miss the corner points or find a false corner points.So it is not very suitable for embedded systems to process image. Phase correlation method utilizes the translation property of Fourier transform. It can overcome the light, uneven gray, fuzzy image.And DSP has a unique speed advantage in Fourier transform processing. So the phase correlation method is best suitable to be transplanted to embedded system.(2) Based on abover, image mosaic algorithm of transplant phase correlation is transplanted into embedded platforms.Matlab environment can not run without PC system. Compared with traditional computers, embedded system has specific application-oriented, system can be cut, longer life cycle and some other advantages.At present, more and more people are starting to focus on embedded system more and more people tend to develop applications in embedded platform. Therefore, my paper transplant the image processing algorithm based on development board Omap3530, set up the development board hardware and software platform, transplant the embedded linux, transplant the image algorithm, complete embedded application from image acquisition to processing, meanwhile, the specific implementation steps and some code are given.