The last step in the development process is to export the images. PhotoLab offers three export formats: JPEG, TIFF and DNG.
Unlike other software, there is no requirement for multiple versions of an image according to its intended use.
The user has to accentuate the image as wish and the system transposes according to the desire dimensions.
– Left Click on one or set of images > “Export to Disk” OR
– Export option list, above right the Image browser
– “Add a new option” button, rename it and activate it (we often forget)
– Quality (JPEG compression index) = 95%
JPEG compression value (and method) differs from software to software. It has no absolute meaning. For PhotoLab, 90% output produces files that are four times smaller than uncompressed files.The 97% value for files half the size is the compromise that works for me.
An interesting option of PL is to give a destination folder and a suffix to the exported images.
– Destination – Custom folder – path = \3_TVHD or .\3_TVHD or ..\3_TVHD
A “3_TVHD” folder will be automatically created at the first export from the current folder.
The path meets unix standards .\ and ..\ (see Table 1).
– Advanced settings – Suffix = _DxO
The suffix is optional but applied permanently It is possible to easily vary the suffix at each export, but we can not to rename it on the fly.
Resize the image
PhotoLab offers six resizing options.
The simplest example is to take the preparation for projection on a 1920 x 1080 pixel HD TV screen of a report composed of Horizontal and Vertical images (landscape and portrait oriented) from a 4/3 format camera of 4600 x 3450 pixels.
Required target image are: width = 1920 x height = 1080.
* Fit is the classic option that all software offers. Its name does not specify it, but depending on its original orientation, Horizontal or Vertical, the output rectangle is rotated by 90° to exploit the largest possible size of the container (the output image).
– Horizontal image is output in 1440 x 1080
– Vertical image is output in 1080 x 1440
The order of the boxes 1920 < x > 1080 is indifferent.
* Rotate to fit, is not understood by the PhotoLab testing community.
According to DxO, this option allows you to rotate the original image to best match the orientation of the output … what the “Adjusted” option already does.
* Longest side or Shortest side.
– The Horizontal or Vertical image comes out respecting the instruction
The dimension of the other side has no constraint
* Megapixel. This is the size of the image in pixels, not the file size 😉
* Proportional. 100% is equivalent to the “Full JPEG” profile option
It is not possible to impose an output file size on PhotoLab.
Resizing in centimeters or inches is the only one that requires that you enter the resolution (in Dots Per Inch or Dots Per Cm).
I created several profiles, mainly JPEG, depending on the destination of the images.
– Full JPEG = full format for comparison of different versions and prints
– HDTV = Rotate to fit 1920 x 1080 to speed up the display on the TV
– Diffusion = Longest side 1280 pixels for a display without rulers on “all” screens, computers and tablets (dimension maybe to be updated today!?)
– Web = Long side 800 pixels compressed at 90% for social networks and mailings
– 16-bit TIFF for post-processing in image processing software (Affinity Photo, PhotoShop, GIMP, …) as well as fine art prints (see table 2)
It is possible to export several formats into a command (Elite edition).
– Validate multiple check marks
Downsizing is more effective at reducing file size and qualitatively better than high compression for broadcast images.
There are three methods of interpolation when changing formats.
DxO advises to export to small images with the bicubic option.
For the record, interpolation in photography was originally a method of producing missing pixels when increasing the image definition. Today it is used to reduce this definition by producing a new pixel from the analysis of its neighbors on the original image.
– Bilinear interpolation is a method that harmonizes the pixels by averaging the values of colors of four neighboring pixels. The calculation is fast but produces results of average quality with visible pixelation.
– Bicubic interpolation is a more accurate method based on the examination of the sixteen surrounding pixels. It produces smooth edges with less pixelation and fewer artifacts but sometimes making the image bland
– Bicubic sharper interpolation is a variant based on bicubic interpolation with improved sharpness by enhancing contrast. This method preserves the details in a resampled image of small dimensions intended for the web.
I did tests on detailed and smooth images (gradients) of 4000 pixels (starting format ratio 1:1).- For an output format reduced by half, I do not see any difference between bilinear and bicubic, and bicubic sharper produces an image that is too accentuated
– With an aspect ratio reduction of 1:4 and 1:8. the bicubic sharper interpolation is interesting
Illustrated explanations here
– Choose a ICC color profile: sRVB, Adobe RVB and others
Mac OS: /Users/~/Library/ColorSync/Profiles/Recommended folder
Prophoto ICC profiles are available from these sites :
(this page also offers a few Prophoto variants)
(don’t mind the ROMM RGB name, it’s Prophoto)
Take a look on this DxO forum page
During installation, PL evaluates the computer performance(*):
* Determines the number of photos to be processed simultaneously when exporting
A inappropriate number of parallel processes, depending on available memory, degrades performance because the system is forced to swap
* Recommends to use the graphics card processor if it is significantly more powerful than the core (Windows)
* The output unit performances (read / write speed, buffer) are essentials
– The Edit / Application – Preferences – Performance menu provides access to these choices
(*) Delete the results file to restart the test.
Windows: C:\ Users \ <user>\ DxO \ DxO PhotoLab <version> \ ocl64.cache
Mac OS: / Users / ~ / Library / Caches / com.dxo.PhotoLab <version> /
TuToDxO.com – All rights reserved – tous droits réservés.