What's involved in resizing and resampling an image

Resizing an image involves redrawing the image using a different number of pixels.

The process of resizing changes the the pixels and necessarily alters the appearance of lines and edges in the image. Consider the case of upsampling a simple 4X4 pixel image to a 5X5 image to get exactly the same image.




It cannot be done. No matter how the new pixels are arranged the the shape changes.

resample40With more pixels, the degree of shape change is not so pronounced but the shape still changes.

The main point is that resizing an image and resampling into a different number of pixels changes the shapes in the image, however subtley, and produces either a softening or jaggy effect on the image which may be more or less pronounced depending how clever the resample41algorithm is that is used to redistribute the pixels.

The mathematics behind effective resampling is complicated. Suffice to say that the 'best' method to use for upsampling is known as 'bicubic smoother', and the best method for downsampling is 'bicubic sharper'.





... back ...