Resize to 30% comparison

Here is a comparative analysis of four methods for resizing a D2x image down to web-size. Three algorithms used in Photoshop, plus a Capture resize, with appropriate USM applied as noted:


Ron

 

I figure it's better to allow you to make your own decisions based on the information presented, rather than telling you... however, I think that the Capture resize is the finest overall (costing a step in post though), with the Bicubic Sharper coming in a close second.

Ron

 

Hi Paul,

That third shot (landscape) is a tricky composition. You happened to have the camera exactly level with regards to the plane of gravity (assuming the buildings in the background were level, as they most likely were), but the buildings are so distant that the viewer's natural horizon is not referenced to them at this size at least. The result is that the entire composition looks rotated counterclockwise. In a large print it would possibly become obvious that you were not rotated (if the buildings were presented large enough to become a visual reference), but in most presentations that shot should have been acquired with a clockwise rotation for a better visual orientation, or rotated in post.

As far as focusing mode selections are concerned (based on Continuous Servo):

Single-Area is for static subjects, or moving subjects which you can keep within the focusing reticle;

Single-Area Dynamic is for subjects which you can acquire within a reticle but you want to be able to track the subject as it moves off the reticle to other areas in the composition (rapidly- and erratically-moving subjects, or moving subjects to be acquired in a fixed composition);

Group Dynamic is for situations where you want to pull a subject or group of subjects out of a featureless (or nearly featureless) field and track them within a small area defined by the group.
(1) is for situations when you have subjects that are moving somewhat erratically, but you can keep them within a small area defined by the group.
(2) is for use when you know they are going to be moving either horizontally or vertically within the defined frame.
Center-acquisition mode is like Single-Area, but with the addition of Dynamic tracking with grouped sensors.
Closest subject will select the closest subject within the group and track it within the selected group of sensors;

Dynamic AF with closest subject priority is essentially idiot mode, but it is useful when acquiring a subject in a low-to-medium contrast scene when you know your subject is going to be the closest strongly-contrasting thing in the frame. This can be a problem in a busy, high-contrast background such as a field of flowers, a water scene with whitecaps, or with a background of trees against the sky where the subject does not have higher contrast than the rest of the scene, or when there are moving subjects in a group (the camera may not be able to select your desired subject), but in some cases it can make the job of acquiring a lock easier (as can Group Dynamic).

Ron

 

Hi Gene,

Open the full-size image (it's a little larger than the visible image, and the resizing algorithm used on this site messes with the detail enough that it obscures information you need to see).

Note the contrast, dimensionality and definition of the eye, beak and breast and wing feathers. Also, check the curved, rising reed in the foreground that is in the focused plane. Note the dimensionality and definition of the head at the transition to the background.

Ron

 

Hi Sandi,

I upsample in steps, but I downsample using integer ratios (80%, 75%, 66%, 50%, 33%, 30%, 25%, or combinations thereof). If I go to 50%, it's in one step. Sometimes, to get a required size it is better to go with a two or more stage reduction (e.g. 80%, 66%, 50%). The idea is that you don't want to attempt to throw away fractional pixels... the algorithm has to invent pixels when you do that.

Ron

 

After the resize... in Capture you use the Size/Resolution dialog (it's not capable of working by percentage, so you type in the correct pixel resize for the percentage), then apply USM as required. Photoshop allows resizing by percentage and algorithm. Do that first, then USM to taste.

Ron