Resolution Conversion Method for Rational Scale Using Neighboring Blocks' DCT Coefficients

Abstract

There are a number of approaches for converting image resolution. In general, resolution conversion takes place in the spatial domain, which is a straightforward method. However, DCT (Discrete Cosine Transform) is often used in image compression. Resolution conversion in the DCT domain is noteworthy because this method can convert image resolution with compressed image data (ex. JPEG and MPEG formats) without decoding so that the resolution conversion process in this method is faster than that in the straightforward method. However, there are many problems in which the scaled up images have blocky noise, the edges are blurry and the basis length after transformation is not the same as before transformation, so general decoders such as JPEG and MPEG cannot be used in this method. We have proposed a new conversion method in the DCT domain using neighboring blocks' DCT coefficients. It is possible to convert image resolution and to improve resolution conversion images by decreasing edge blur and blocky noise. However, using this method, it was not possible to scale up at rational rates, and it was not useful for resolution conversion. So, we propose an expansion to rational rates for this method.