Image Size Conversion Using Two-dimensional Separable Interpolating Filters

With the development of multirate signal processing and digital image, interpolation of images to higher resolution is an important application in the field of image processing. Practical digital imaging applications require the ability to scale images with different resolution at any rational ratio. Image size conversion is virtually an operation that alters the sampling rate of image.In this paper, we research the problem how to scale image at arbitrary rational ratio.An algorithm of designing the anti-aliasing filter and the Mth-band interpolation filter in time domain is firstly provided. The design of filter is translated into an optimization problem with some constraints. Especially, regularity as one of the constraints assure the flatness in passband and the high attenuation in stopband, even the value of stopband response is very close to zero. Several experiments are presented to demonstrate that the proposed decimation filter and Mth-band interpolation filter have good high frequency characteristics.Subsequently, the proposed anti-aliasing filter and interpolated Mth-band filter are applied to the image size alteration and a new approach to rational image scaling called separating variable method is provided. In this paper, two pattern of this separating variable image scaling conversion method are addressed-direct structure and efficient structure. We have presented results that demonstrate that image scaled using the proposed approach can preferably preserve the characteristic of original image. Moreover images scaled using the proposed method are free from artifacts, without any prefiltering, having better visual effect.