Sharpening and Resizing in Photoshop

Resizing reduces the number of pixels in an image. Resizing is done to reduce the size of the image file so that in downloads faster on the internet or will take less time to send as an e-mail attachment. You can store more images on a memory stick, a disk, or a web server if they are somehow made smaller. Sharpening an image is the process of making it appear to have more detail by empahasing the edges in the image.Reducing the resolution tends to reduce the sharpness of the image, so the topics of resizing and sharpening go hand-in-hand. We usually would like to restore some of the sharpness lost by reducing the resolution. You may also want to sharpen an image without resizing it. the sharpening technique is the same.

Image Compression

Save JPG images at level 10 on your computer disk, and use level 7 for e-mail attachments

Another way to reduce the storage required for an image is to compress the image. Rather than store the red, green, and blue values separately for each pixel in an image, a sort of shorthand notation can be used.  More details on compression are in Compress Images.


Here is an image (taken in Fresno, California) which was resized from about 2400 x 3600 pixels to about 320 x 460 with a single Photoshop command.  The command was Image > resize then entering 320 in the dialog box that appeared:

Image resized to 320 pixels high

The image was then sharpened using Enhance > Unsharp Mask... and entering 70% and 1.7 pixels in the dialog box that appears. The threshold is set to 10 steps.
Mask menu arrow Mask dialog

The result is:

Sharpened after one pass resize

The image looks sharper (or at least it should look sharper) because the unsharp-masking process lightens the pixels on the bright side of dark edges and darkens the pizels on the dark side of the edge, so as to empahize the edge. The parameter we set to 70% tells how much to lighten and darken, and the 1.7 pixels indicates how far from the edge to apply the effect. Using 70% and 1.7 pixels is heavy-handed sharpening, but acceptable for images destined for display on the internet. The threshold is the amount of contrast across the edge necessary for triggering the process. It defines how much of an edge counts as an edge. You might use 50% and 1 pixel for an image to be printed.

There is a way to get sharper results that requires more work.  Rather than resizing from 2500 to 320 in one step and then sharpening, go by steps of no more than half.  For our example, we might use resize from 2400 to 1200 > sharpen > resize from 1200 to 600 > sharpen > resize from 600 to 320 > sharpen.

Using the sequence of steps yields:

Resized by steps

That looks sharper than using one reduction plus sharpening.  We can compare an enlargement of a small area near the center of the two images to better see what happened. The left image is from one step, the right image from the multiple steps:

Detail after one pass resize  Detail after Step Resizing

If sharpness is important, the process with multiple steps is much better. You have to decide in each case whether it is worth the effort. If you have a copy of one of the expensive versions of Photopshop, such as CS2 or CS3, it is possible to define a script that encapsulates the sequence of steps into a single command. The end result is no better than doing it manually, but getting there is easier once you have defined the script.

Why does working in steps produce better results?  Resizing reduces some number of pixels into one.  The default method in Photoshop is the best method they offer, called bicubic interpolation.  It considers three pixels horizontally and three pixles vertically to compute one output pixel. The input pixels nearest the one out are weighted the most. To get from 2400 x 3600 to 320 x 460, a block of very roughly 8 x 8 pixels, 64 in total, around each output pixel might be considered  for contributions to the computed output.  If the reduction is done in a single step, only six of the64 pixels in the large block will be used to compute the one output.  However, if we go in steps the 8 x 8 will first be reduced to 4 x 4 with all of the original block contributing to that reduction. Then all of the pixels in the 4 x 4 will be used, and so forth.  The final result is better, because 64 rather than 6 pixels in the original contributed to each pixel in the reduced image.

If the sharpening is too strong on each level, the result may show stairstepped computer artifacts. Our 1.7 pixel and 70% is at least on the verge of that, if not over the top.