With the concept of information super-highway was proposed, in- ternet is widely used, the high clear (high resolution) panel TV, the high definition DVD and projector are popular, the human being have been in digital home hd era. The emerging industry-digital home should be free to timeliness, motility and cost performance, it is a hot case in current manufacture as well as in research eld, there are a lot of challenging works to process and analyse hd media. Hd media actually is composed by many frames of images, so we can classify video processing as image processing. Many digital image processing software generally resize image by polynomial interpolation,but the results are not very ideal. A bunch of researches show that we can get more ideal results via nonlinear rational function models. Bivariate blending continued fractions rational interpolation is not only a very efficient nonlinear approximate, but also can be calculated recurrently, which means good feasible computability. This paper takes Bivariate blending continued fractions rational interpolation as theoretical basis, generalizes Bivariate blending rational interpolation on regular mesh as Newton-Thiele type continued fraction on irregular mesh, and combines with digital image feature, applies it in resizing image to get better results.