Image resizing and rotation based on the consistent resampling theory

The recently proposed consistent resampling theory for non-bandlimited signals is applied to image resizing and rotation. Images with high frequency components can be resampled using this scheme to achieve high quality performance. Image resizing is treated as resampling using non-ideal interpolation functions. Both zoom in and zoom out by non-integer factors are considered. Image rotation is also formulated as a resampling process. We show that our approach outperforms other linear image processing techniques without increasing the computational cost.