Interpolation And Background-Color-Clearing Method Research For Engineering Scanner Software

Large scale high precision scanner is an indispensable image acquisition device. It is also a key device in the procedure of drawing-digitalization. Scanner system not only needs high quality hardware, but also needs excellent image-processing software. In this paper, Firstly, two important implementation strategies during the scanner software development are discussed, including effective management of massive image data and easily undo. Secondly, a detailed introduction of three kind of interpolation methods are made, including Nearest-Neighbor, Bilinear and Cubic Convolution, with the derivation of the interpolation equations, the experimental data shown that these methods are available. It is necessary to clear background-color offset before binarization because the most drawing papers have asymmetric discolorations. An algorithm based on mathematical morphology which is used to get the available image information are used in this paper, another method based on the fourth central moment criterion which is used to segment the image is introduced. The experiments are shown that the method is available for some kind of image but not available for others. It needs to be improved in the future. Finally, some principles of code-optimization are described, the MMX instruction set are used to improve the performance of bilinear interpolation code. The test result is shown that the method based on SIMD instruction set is very useful in image processing. We must pay more attention to it.