An example of Resampling is shown by the giraffe photos. The original image was taken on a 6 megapixel camera with an image size of 3072 (height) by 2048 (width) pixels, but later cropped to 1985 by 2048 pixels. To put the cropped image onto this webpage, it had to be reduced in size as the current dimensions were larger than a typical computer monitor of 1024 by 768 pixels. What I wanted was an image about 320 pixels on the longest edge. If I had merely cropped the original image to the required size, it would have fitted on a webpage, but instead of seeing the whole of the giraffes head (right) only its eye would be seen (below right). Both images are approximately the same size in terms of dimensions, but the end effect is completely different. Essentially, resampling an image is the process of physically changing the dimensions (pixel size) of an existing image file.

Avoiding Confusion
Before progressing further, it should be noted that references to file size in this tutorial concerns the physical dimensions (height and width, in pixels) of an image, and not its memory size (megabytes or kilobytes). This is an area that often causes confusion for newcomers to digital photography.

Effect of Resampling on Image Quality
Although resampling can be used to increase or decrease an images dimensions, there is always some loss of quality involved in the process, and quality probably suffers more when increasing file sizes, particularly where increases are large (eg, more than 100%). If your main requirement is to increase file size by large amounts, it may be worth investing in specialist software such as Genuine Fractals, otherwise, PhotoImpact or similar image editors should generally be sufficient for resampling work. Providing that the original image is generally good quality (ie reasonably sharp), and the change in file dimensions following resampling are not excessive, then the quality of the final file (or print) should be good.
The following parts of this tutorial show how to resize images using PhotoImpacts Resize dialog box.

Step 1: Open the Resize Dialog Box To open the Resize dialog box, click Adjust on the Main menu bar, and select Resize. The dialog box shown right will appear. There are a number of settings to consider and these are discussed in the sections below.

The Active Image section shows the current image size, in pixels, and the current resolution, expressed as the number of pixels per inch (ppi). Digital cameras often produce jpeg files with a resolution of 72 ppi, so its not uncommon to find this is the default resolution of an image when opened in image editing software such as PhotoImpact.
If the resized image will only be displayed on a computer monitor, there is no need to change the resolution as 72ppi is fine for this purpose. If the resized image is for producing a printed version of the file, then a higher resolution such as 300 ppi would produce a better result. More information is given on this in the 'Resolution' section below.

Step 2: Identify what needs to be resized This section is used to identify which part(s) of the current image will be resized. Often, it will be the 'base' image, or a copy of it, but for images that have been edited and contain 'objects' - known as 'layers' in most other image editors, there is also the option to resize either a selected object (layer), or the base image and any objects it contains as well. Unless your image contains any objects, the Base image will be the appropriate setting for resizing an image.

Step 3: Identify the required output sizeThe next step is to identify the required output size of the image. These can either be Standard (default) or User Defined image sizes. Before selecting a size though, ensure the 'Keep Aspect Ratio' box is checked (highlighted red, right). This keeps the height and width ratio of resized images the same as the original source file, otherwise you may end up with a distorted image.

Select the preferred measurement unit from the picklist (highlighted blue, right). In this example, 'Pixels' have been selected. Click the 'User Defined' checkbox and enter the required size for the output file. If the 'Keep Aspect Ratio' box is checked, you only need to enter either the Width or Height as the second size will be calculated automatically from the first entry.

You may wish to explore some of the Standard sizes available on the picklist, but ensure that the height/width proportions of these are identical to the height/width proprtions of your image otherwise the resampled output file may be distorted. If in doubt, always select the User Defined option, and ensure the Keep Aspect Ratio box is checked.

Adding User Defined Sizes
Its possible to add your own customised sizes to the User Defined options. This is useful if your original source and resampled output files are aways the same height/width proportions.

To add a customised size, click the red tick icon (highlighted red, left), type in a relevant name and click OK. The name you add will be displayed on the menu for the user defined size options. To access the size you defined, click the red tick icon which will display your custom file size on the pick list options (see below).

Step 4: Set the Resolution If the resized file is intened for use on a webpage, or for emailing to a friend, leave the resolution at 72ppi as this is the default setting for webpage images.

If the resized file is to be used for printing, then a higher resolution is required. For best quality printing, set file resolution to around 200 to 300 ppi. Its said that the human eye is unable to resolve resolutions of more than 300 ppi, so anything higher than this produces unnecessarily large file sizes.

Approximate output print sizes for combinations of image size and file resolution are shown in the table below. To calculate the print size for other combinations , simply divide the image dimensions (pixels) by the file resolution (pixels). This will give the output print size in inches.

Controlling Print Size by Changing File Resolution The table shown above illustrates that print size can be controlled, to a certain extent, purely by changing the file resolution. In other words, its not always necessary to resample a file to produce a certain output size. For example, if you have an image file with dimensions of 2000x1500 pixels, and want to produce a print of approximately 5x4 inches, setting the file resolution to 400ppi will produce the required print size (ie 2000/400=5) without resampling the file. This is a better option than resampling as the original file dimensions are left unchanged, which means there is no loss in image quality. By contrast, resampling a file always involves some loss of image quality.

This method does not work well if the file resolution required to produce the print size is lower than, for example, 100ppi. Using a low file resolution will probably produce a print with pixels that are too prominent to produce continuous tone as normally seen in a photo. The image shown right illustrates what a print with a low file resolution would probably look like.

Step 5: Select the Resample Method PhotoImpact has three ways of resampling images, namely Bicubic, Bilinear and Nearest Neighbour. These are essentially mathematical equations, known as interpolation algorithmns, which add new pixels based on the existing pixels in an image. Fortunately, its not necessary to understand how they work, but it is useful to know about the characteristics of their output to decide which method to use. The Bicubic method produces a soft image (which can be sharpened) with the smallest file size, while Nearest Neighbour produces a sharp image with a larger file size. Bilinear resampling produces an image approximately midway between the other two methods in terms of sharpness, but with a relatively large file size (ie, in terms of megabytes or kilobytes).

Select the preferred resampling method from the picklist and click OK. The file will then be resampled (resized) according to the options selected in the Resize dialog box.

When editing images you may wish to consider using a duplicate copy rather than an original, to ensure your original file is unaffected if any problems occur during the editing process. This is good practice, and a habit worth developing.

Comparison of Resampled Output Files A comparison of files produced using PhotoImpacts resampling methods are shown below. Because the images were saved as jpeg files to display on this webpage, quality has suffered a little due to file compression artefacts, but even allowing for this, subtle differences can be seen between the different methods. The methods shown are bicubic (top left), bilinear (top right) and nearest neigbour (bottom left).

The preferred method for image resampling will of course vary with personal preferences and quality of the original images. In general, I use Bicubic resampling most of the time, and although this produces a softer image than other methods, tonal quality is generally better and sharpness can be increased with use of the Unsharp Mask tool. Bicubic resampling also has the advantage of producing smaller file sizes (ie megabytes) than the other two methods. The final image (below) was resized using bicubic resampling, then sharpened with PhotoImpact's Unsharp Mask tool.