This topic came up over at CANP, so I thought I'd pose the question here and request help from some of the gurus...
Regarding incremental resizing, i.e resizing downward in 50% increments: I would like to know how well it works. Is there less noise? What exactly does incremental resizing do for the image?
And while we're at it...
Which algorithm is best: Nearest Neighbor; Bilinear; Bicubic, Bicubic Smoother; Bicubic Sharper? Does the method used depend upon the specific image? I've always found this to be confusing. I usually use Bicubic, but if another method works better, I'm all over it.
Use bicubic smoother for upsizing, it is a modified version on the bicubic algorithm designed to behave as if we upsized in steps.
Use bicubic sharper for downsizing.
The theory behind the incremental resizing is that by taking smaller steps, the algorithms have an easier job generating the new pixeles and therefore generate less jagged edges and smoother transitions.
The downsizing version also included a small sharpening step between each reduction in size, this was designed to help preserve detail that might otherwise be lost as the calculations discarded information.
Posted by Bob Steele on January 1, 2007 at 8:42 PM
Bicubic smoother for upsizing, standard Bicubic for downsizing (all in one shot).
Preston: I use only bicubic in one fell swoop, but I'd suggest you run your own tests to determine your results. It is recommended by Adobe that one use Bicubic Sharper for downsampling but I don't like the BS results compared to bicubic. I'd rather do my own sharpening as the final step before printing.
Bob, do you prefer to downsample your originally sized images in one shot for the web usage? I tried to compare step-by-step and single step method and saw no differences between them indeed.
Posted by Bob Steele on January 3, 2007 at 9:17 AM
Bob, do you prefer to downsample your originally sized images in one shot for the web usage?
Yes, for all downsizing applications, including web sized images - using standard Bicubic interpolation. Then apply sharpening appropriate to the final size of the image (web, print, etc.).
Preston, there is an excellent article on Ron Bigelows site that covers this topic..
http://www.ronbigelow.com/articles/interpolation/interpolation.htm
I have used Bicubic all along and then do my sharpening as a last step. It appears to work very well for me.
Bob: There can never be too many birds. I'm sure glad that there are more than enough for you photograph! The same goes for you other wingnuts! Your skills at this endeavor always amaze me.
Thanks again!
-PB
Posted by Chris Ross on January 16, 2007 at 11:45 PM
I started using a routine for downsizing for displaying images here on NPN a while back, it doesn't take a great deal of extra time so I use it for all images, though it probably only makes a difference on some images. I started using it because I was having trouble resizing images like forest scenes with a lot of fine detail all over the frame.
Assuming I am aiming for a 640 pixel image on the long side I resize the long side to 2560 pixels with bicubic (or I now tend to use bicubic smoother). Then I go to 1280 then 640 pixels with bilinear, on the basis that he calcs are easier when exactly halving pixel numbers. I wouldn't necessarily suggest using it on all images, but could be worth a try on those very detailed images that don't JPEG well. Easy enough to form into an action too.
To Kevin,
Thanks for the useful link. I actually just skipped to the updated article "Interpolation Revisited" and it very nicely explained what I needed to know.
Pete D.
---------------------------------------------------------------------------------------------
Reply posted by Kevin Thompson on 01/03/07 at 11:33 am EST
Registered on 03/15/05, 4 Topics, 28 Replies
Preston, there is an excellent article on Ron Bigelows site that covers this topic..
http://www.ronbigelow.com/articles/interpolation/interpolation.htm
Discussion
10 repliesRegarding incremental resizing, i.e resizing downward in 50% increments: I would like to know how well it works. Is there less noise? What exactly does incremental resizing do for the image?
And while we're at it...
Which algorithm is best: Nearest Neighbor; Bilinear; Bicubic, Bicubic Smoother; Bicubic Sharper? Does the method used depend upon the specific image? I've always found this to be confusing. I usually use Bicubic, but if another method works better, I'm all over it.
btw: I am using PS2.
Thanks for your help!
-PB
Use bicubic sharper for downsizing.
The theory behind the incremental resizing is that by taking smaller steps, the algorithms have an easier job generating the new pixeles and therefore generate less jagged edges and smoother transitions.
The downsizing version also included a small sharpening step between each reduction in size, this was designed to help preserve detail that might otherwise be lost as the calculations discarded information.
Yes, for all downsizing applications, including web sized images - using standard Bicubic interpolation. Then apply sharpening appropriate to the final size of the image (web, print, etc.).
http://www.ronbigelow.com/articles/interpolation/interpolation.htm
I have used Bicubic all along and then do my sharpening as a last step. It appears to work very well for me.
Bob: There can never be too many birds.
Thanks again!
-PB
Assuming I am aiming for a 640 pixel image on the long side I resize the long side to 2560 pixels with bicubic (or I now tend to use bicubic smoother). Then I go to 1280 then 640 pixels with bilinear, on the basis that he calcs are easier when exactly halving pixel numbers. I wouldn't necessarily suggest using it on all images, but could be worth a try on those very detailed images that don't JPEG well. Easy enough to form into an action too.
Thanks for the useful link. I actually just skipped to the updated article "Interpolation Revisited" and it very nicely explained what I needed to know.
Pete D.
---------------------------------------------------------------------------------------------
Reply posted by Kevin Thompson on 01/03/07 at 11:33 am EST
Registered on 03/15/05, 4 Topics, 28 Replies
Preston, there is an excellent article on Ron Bigelows site that covers this topic..
http://www.ronbigelow.com/articles/interpolation/interpolation.htm