The Image Interpolation Algorithm Based On The Non-uniform B-spline Surface

In our daily lives and work, we often need to change the size of the image to meet the needs of different areas and applications. Digital Image Scaling has become a very important topic in the field of digital image processing. The image interpolation is an important tool for image scaling technology. At present several commonly used algorithms include the nearest point interpolation, bilinear interpolation, the bi-cubic interpolation as well as the spline-based interpolations. These algorithms are classified as the linear interpolation technology. Their interpolation principles are all similar to the "low-pass filter," which bring degradation to the details of the original image in some ways. In recent years, some new algorithms for image interpolation as methods based on WT, methods based on partial differential equations, artificial neural networks and the Bayesian analysis, etc. have their advantages during interpolation, but it is still difficult for them to avoid the drawbacks of loss of edge details in interpolated results.Based on above background, a new interpolation algorithm based on the non-uniform B-spline surface is proposed. In this paper, the main research contents contains: the methods of parameter selection when constructing B-spline curves; the construction of the B-spline surface, as well as how to deal with artificial edgs in the interpolated image. There are some main innovation points in this paper as follows:1. It applies a non-uniform parameterization method based on the change of the brightness of pixels to construct the B-spline surface. The whole image is put into the three-dimensional space. The values of image pixels are taken as the values of Z axis in three-dimensional space. It applies a non-uniform parameterization method based on the change of the brightness of pixels to construct the B-spline surface, which contains grids computed with each 25 adjacent pixels in the image. It is continuous with C1 level. By this way the objective image obtains a higher resolution and less noise than the traditional spline interpolations.2. We extract the edge pixels from the image in the way of edge detection.For the serrated phenomenon on image edges caused by above algorithm, we will separate the edge pixels from the image by edge detection.3. We interpolate the edge pixels with the bi-cubic algorithm.There are noises in the image after interpolated by bi-cubic interpolation algorithm, but good visual results on the edge can be obtained. Therefore we interpolate the image edge pixels with the bi-cubic algorithm while the other pixels will be interpolated by above algorithm based on the non-uniform B-spline surface. Thus, the fuzzy or jagged phenomenon is weakened on edges when the image is scaled to larger size. The whole image quality can be ensured.In short, in this paper, we combine computer graphics and image processing theory and then propose a model suit for the scale based on any scaling. It provides a new perspective for the field of digital image processing. The results have a good theoretical significance and practical value in the fields of medical, police, military, weather and so on.