I appreciate your receptiveness. After a good nights rest, a quick one.
I hope its ok to be open, and state that as part of any tests or review, I am obviously comparing Filmulator with other tools. Winter is only winter, because we have other seasons like summer.
I noticed that the pixel size changes from the original.
A original raw file captured in Sony Nex-F3 digital camera as a 4912 x 3264 image, becomes saved by Filmulator, in both tiff and jpg formats as a 4920 x 3276 pixel image.
But Filmulator is not alone in this “slight” alteration of the aspect ratio.
I had always felt that there was something not quite right about the portraits which I processed in darktable, and now I can clearly see why, from an objective opinion, the same image, and in fact all my images from this camera are saved by darktable as 4928 x 3276.
The in-camera jpgs are exactly the same pixel width and height as the raw image. And when I check the output of Capture One 21, which is free to download (but needs logging in to get an account/license) it also generates images in jpg at the same size as the raw image.
All knowledge is positive. But this makes it more difficult to directly appreciate any differences between the “colour science” of the different apps, cos as you switch from one image to another in the image viewer, you eyes also have to adjust to this instantly modified aspect ratio.
This is a pretty interesting observation, and in all the brief discussions I have been priviledged to read online, or watch on youtube, this issue has not ben raised with respect to darktable, which has already a good number of users.
When I go into the various photo raw tools, and check the image proprties in the tools, each of them displays a corresponding set of width and height pixels that corresponds to their jpg and tiff outputs/exports. So there is consistency between what the app “thinks” and what it generates, but there is no consistency between the apps, and the two open source apps that I have checked - Filmulator and Darktable, deviate, in different ways, from the original image pixel size, which is adhered to and not altered by the three other apps I have tested with - Adobe Photoshop Express, Capture One Express for Sony, and Sony’s own Imaging Edge app - all these closed source apps are free to download.
Clearly as the open source photo apps rely on some common libraries which are most likely not created by the developers, but reused, I deduce that the this deviation is coming from the image processing libraries, e.g libraw or rawspeed.
Cosmetically - not a big deal, but if one had a forensic kind of use case, or different human editors working on a large pool of images taken on the same camera, but using different photo editing tools, this slightly varying pixel size results, would be something one would wish to avoid.
I appreciate that it may not be something you can do anything about, cos you most likely did not develop all of the raw-image library which you use, but it gets even more interesting.
I also checked with Raw Therapee only because it is supposed to use the same image processing library for extracting the image from the raw file, as Filmulator, and in Raw Therapee it gets it absolutely right, and maintains the identical pixel size as the original raw file. So it is possible for an open source app to get this absolutely spot on.
But we go one step further, cos when I look at the exif data via Raw Therapee this deviates from the original size to 4928 x 3276 neverthelesss Raw Therapee still correctly displays in the image viewer, and generates jpgs in the original pizel size of the original raw file, using some other method (could be there are two aspect definitions in the exit data - I have not checked the exif using other tools) or a companion database of “corrections” which RawTherapee uses. !!
So we use an Exif viewer and the reasons for some of the above, become a bit clearer or easier to understand.
I used Exif Tool Gui, which relies behind the scenes on some other open source console app. Much easier to use a GUI.
In the exif, we have different values.
- ImageWidth 4928
- ImageHeight 3276
then we have another set of values
- SonyImageWidth 4912
- SonyImageHeight 3264
So it appears that the correct info is there, but each app may have a choice in how it decides to use their image raw library to determine the size of the image.
From an accuracy prspective, I do not think there should be any discrepancy or difference between apps, on something as objectively accurate as the pixel height and width. And if the app were to deviate from the exact size of the original raw image, for whatever reasons, it would be good to let the user know or give them a choice.
Especially when taking photos of products, and people, you definitely want to get the aspect ratio absolutely right, otherwise, you have inadvertently distorted their image, irrepairably.
I tell you no lie, a few weeks ago, I was uncomfortable about the image of my wife, whom I obviously know well enough, as I processed some portrait shots in darktable, just did not look right, now I can appreciate one good reason for this. The image had been “distorted” and I had absolutely no clue that this had been done, cos I trusted the app.
I am sure this “bug” is something you can fix in the next release. Best wishes.
I have done, within a few minutes, this morning an extensive comparison of the images generated by different image processors, but this will need a thread of its own, and it may take me about a month or two to find the time to present my thoughts, in a way that allows others to make their own objective and subjective conclusions. That should be another pretty “revealing” thread. As we are speaking of images, the pun(joke) on revealing is intentional.
I hope everyone would agree with me, imaging apps, should not make their own mind up about the pixel size of the image, without the permission or consent of or explicit notification to the end user…! Hopefully unlike the subjective issue of color science, we can all objectively agree that pixels should be presented with the utmost accuracy size wise, especially in a digital domain where this accurate information is available in the raw file.