Improvements On Bicubic Image Interpolation

Bicubic algorithm has the advantages of more accurate image magnification and faster processing speed. Nowadays, bicubic amplification technology is adopted in many image processing software, such as Adobe, printer driver, etc. In this paper, an improved bicubic interpolation algorithm is proposed. Unlike the traditional algorithm, which sets the key parameter a to - 0.5 in order to guarantee Taylor's third order approximation, we remove such requirement because this value is only applicable to the interpolation of flat area while the change of image pixel value is often irregular By directly optimizing the mean square deviation of the image, we obtain the higher-order optimization function. At the same time, a new variable is introduced to replace the a
²
term of the optimization function, so that the higher-order optimization problem can be transformed into the least squares problem. Experiments show that the optimized parameters have better reconstruction effect, and the acquisition of the optimized parameters is directly related to the magnification of the image, and the image itself has little influence.