The Study And Implementation Of Geometrical-Distortion Auto-Correcting Algorithm Of Digital Image

According to the problem that geometrical distortion occurs in the images taken by digital camera easily, this paper studies and realizes a geometrical-distortion auto-correcting algorithm, which is valuable both theoretically and practically.In the research, careful and profound analyses are made to the practical situation in which the geometrical distortion occurs, and consequently, a method to auto-correct the geometrical distortion is educed. This paper puts forward a geometrical-distortion auto-correcting algorithm basing on auto-extracting boundary and corresponding-tiepoints geometrical transform, and uses 2 kinds of detailed method to the building distortion and business card distortion. In the program of building, first use the histogram equalization to enhance the image, use Sobel operator to detect edge, then use Radon transform to auto-extract the boundary and at last correct the image. In the program of business card, first use Otsu algorithm to do image segmentation, get an image in which the business card or paper is separated from the background by region filling and boundary extracting, then use Radon transform to extract the boundary; when calculate the corresponding tiepoints, put forward the image mathematic model of digital camera, successfully calculate the aspect ratio of business card or paper to make sure that the business card or paper are not distorted; after the first correction, use Numeral Character Recognition Technology to judge whether the image needs to be overturned.Finally, the algorithm has been proved by experiments that this restoration method has good performance and gets the one upon when comparing with other similar software. It is appraised well by PENTAX and the embedded work to the digital camera has been finished.