Study On Pixel Compensation Technology Of Digital Image Based On Sub-pixel

The Digital Image Processing system combines optics, sensors, image processing andpattern recognition and some other critical technology together, enabling the quickmeasurement of objects’ size or relative position. The system is a noncontact measurementsystem based on image processing, which takes digital image and inputs that into computerby using high precision CCD Cameras, then digital image is converted into digitalinformation and analyzed and processed by the software, finally the geometric parametersof digital image was obtained. In the transportation and processing process of digital image,image information will be changed due to noise and distortion, then error will happen. As aresult, the data will not be as good as that from ideal condition. So the compensation ofimage distortion is the most important.A digital image measurement system is descript by this paper, which creates thefoundation of precious measurement. According to the structured edge features, smoothimages and edge gradient are guaranteed by using Anisotropic Diffusion Filter AlgorithmProcessing. The conflict between smooth image and edge stability is well solved by thismethod. First of all, single pixel accurate edge is obtained after rough location with Cannyprocessing measure. Then sub pixel is preciously located by the application of QuadraticPolynomial Interpolation Algorithm, based on step edge features.The accuracy of measurement is guaranteed effectively by introducing comprehensivesystem calibration process which is based on linear imaging features. The method isproposed to describe the ideal image point position and the actual one‘s Dual three timeserror compensation function by the extraction of the straight edge of1gauge block. Opticsdistortion error, perspective error, sensor positioning, positioning error of edge detectAlgorithm and other systematic error can be solved comprehensively by this method.Several times measurement of different gauge blocks of the system will be done, and a±10μm measurement accuracy is proved to be achieved by this method.