Correction #1 - BlurA word about blur, sharpness, sharpening and workflowThere are two types of sharpness and it is important not to confuse them. Optical sharpness is determined by the quality of the lens and the sensor. Software sharpness creates an "optical illusion" of sharpness by making the edges more contrasty, compensating for some of the softness caused by sensors with a color filter array. Software is of course unable to create detail beyond the camera's resolution, it will only help to bring out captured detail by increasing the perceived sharpness. Software cannot create detail that was not originally captured by sensor.
If you are shooting in RAW, you can "undo" the in-camera sharpening. If you are shooting in JPEG, I recommend you shoot with in-camera sharpening set to "LOW". As we will see in a moment, it is not always straightforward to sharpen JPEG images shot with sharpening "OFF" so that they look as good as those shot with sharpening set to "LOW". The "NORMAL" and "HIGH" settings tend to create difficult to remove sharpening halos. Low in-camera SharpeningDxO Optics Pro definitely improved image sharpness. However, in this image, this was very similar to a simple Photoshop "Unsharp Mask" with Amount 50%, Radius 1.7 (not my usual preference) and Threshold of 1. The minor differences in the roof tiles are nearly invisible at 100%. It is very easy to create a Photoshop action if you want to apply this type of sharpening to all your images. The 50 x 50 pixel crops below are 2 times (pixel) enlarged via nearest neighbor interpolation.
No In-camera SharpeningGiven the above result, I dug a little deeper and tested DxO on images with in-camera sharpening set to OFF. As stated in my sharpening article, sharpening images without in-camera sharpening can be tricky. DxO achieved an excellent result, somewhere between LOW and NORMAL in-camera sharpening. In order to determine whether Photoshop could achieve a similar result on this D70 image, I first applied an Unsharp Mask (Amount 100%, Radius 0.9, Threshold 0) followed by a Gaussian Blur with Radius 0.3 to smoothen out any artifacts. This first step ensures there is enough remaining sharpness after the Gaussian Blur. I then applied a second Unsharp Mask (Amount 50%, Radius 1.5, Threshold 0). These parameters were chosen not because they are my usual preferences, but in order to achieve a result similar to DxO for the purpose of this test. The results are very similar in different areas of the image.
Normal or Higher In-camera SharpeningOnce the image is already sufficiently sharp, it makes less sense to apply the DxO blur correction as that will increase sharpening halos (especially visible in Crop 5 below). The result is very similar to a Photoshop Unsharp Mask with the settings of Amount 50%, Radius 1.7 (not my preference), and Threshold 1. Again, if you want to apply this type of sharpening, it should only be as a final step. Also note that even if your in-camera sharpening is set to the highest level (not recommended), DxO will continue to sharpen the image, leading to very visible sharpening halos.
CompressionObviously you would normally not use a D70 to shoot in JPEG BASIC mode as that makes the JPEG squares visible. The example below is just to illustrate the danger of sharpening JPEG images as it enhances the visibility of the 8x8 JPEG squares as shown in the "After DxO Optics Pro" crops. A smoothening of the JPEG squares before sharpening would have been great.
RAWThe first two columns repeat the results of the Low and No in-camera sharpening results. The last column shows that RAW is really the way to go. Photoshop extracts virtually all detail captured by the camera, very similar to what Nikon Capture 4.1 achieves, and better than any of the other DxO and non-DxO examples. Viewed at 200%, the roof tiles in the RAW image could use some anti-aliasing, but look good at 100%. And there is more, see below.
* (Foliage moved between shots) The crops below shows that if you are really serious about extracting detail, RAW is the way to go. The loss of detail caused by the maze artifacts was not recovered by DxO Optics Pro (it would not be possible on a JPEG), but very well recovered (*) by Nikon Capture 4.1 which used the RAW data:
(*) the color moire is also eliminated, while a less visible moire pattern is introduced. NoiseWhen sharpening, one must avoid applying the sharpening to uniform areas as that will only enhance artifacts. DxO claims to sharpen less in uniform surfaces. However, this appears to be not much different from the fact that an Unsharp Mask does that by default anyway, especially if a certain Threshold is chosen. More sophisticated localized sharpening can easily be achieved by sharpening a duplicate layer with a mask. Corner AreasTo finalize my quest to find some rocket science behind DxO Blur I did an analysis on whether there was any differential sharpening in the (typically softer) corner areas caused by the lens (e.g. crop 1) at different apertures. But even at 4 times enlargements I could not detect anything statistically significant. Conclusion on the DxO Blur CorrectionDxO Optics Pro certainly gives an improvement in terms of perceived sharpness and does so in an automated way. Unfortunately it is only possible on JPEG images. Also, the lack of user controls will lead to sharpening halos if the in-camera sharpening was set to NORMAL or higher. But more importantly, even if DxO did some type of "intelligent sharpening" with vastly different results from a conventional unsharp mask, this benefit would be lost to a great extent by the fact that DxO requires this to happen at the beginning of the workflow before noise reduction, contrast adjustments, resampling, etc. We will now review the other three DxO Optics Pro corrections.
|
| Bookmark: |  Digg |
 del.icio.us |
 Reddit |
 Google |
 |
||||
| Actions: | < Previous | Next > | Print page |
