The “after hours” thing is likely a good way to go if you have that option. Unfortunately, I don’t.
I also wanted to let you guys know that I found another stitching utility that seems to work even better than Stitch Panorama. It is a stand-alone utility called Hugin. It is primarily designed to build a panorama from digital photos, taking into account lens types and settings, but there is a tutorial on how it can also be used to stitch scanned image files. It does a very good job and it’s also very fast.
Thank you for letting us know about both. I always used to do this by hand in the GIMP, and that’s a lot of work, and I have to do colour balance myself.
Yes, 600dpi should be enough to capture the print screen accurately. Lower resolution may cause moire patterns, higher is hardly an improvement.
If you don’t want to bother with descreening your images (which can be a pain if you want to do it properly) it’s probably a good idea not to scale them down. If you must, make sure to use a resampling algorithm that doesn’t bring out the moire so much.
Hugin is great. I use it to stitch partial scans of 12’’ record sleeves.
I wouldn’t recommend picking black and white points. It will usually clip shadows and highlights. Picking a neutral point if necessary works a lot better for me.
If you really want to get accurate colors you could get a software like Silverfast to create a calibrated color profile for your scanner, but it sure ain’t no budget solution…
I usually just scan in 1200 DPI to PNG and (after maybe doing a slight bit of cropping) upload that. If people want more polished cover art images, I have posted my source material and others are free to download and edit/modify/re-upload. (And @rdswift says he’s lazy? Psh. )
Note that CAA will “soon” have a new max-dimension=1200px download option (right now the biggest that isn’t “original size” is max-dimension=500px):
so your 1000×1000 images will actually be smaller than this.
Also, do you square cover art even for non-square covers (e.g., digipaks)? (I hope not!)
Then it’s probably a good thing that I just this morning decided to change to 1500 x 1500. Actually, with the way things are changing, I think I may change that to 2000 x 2000 px.
I do, but I also upload scans of the full digipak covers so it’s all there for the archive. My reason is that, unless you specify in Picard to use the original image size rather than 500 x 500 px (for example), MB / CAA / Picard will take the largest dimension of the file and resize it to the specified size, and then fill in the smaller dimension with a blank background to create a square image. I would rather have a nicely cropped square image than one with blank filler bars on the top and bottom.
On those rare occasions when you just can’t crop an image enough to make it square, I will create my own square image with an appropriate color / background used for the filler bars. I don’t upload these, but just use them for tagging my own collection.
I’ll try the neutral point next time I’m working on cover art, but I’m not talking about taking some random point as black. My scanner shows them both as somewhat grey, and all I’'m doing is restoring the parts of the image that should be completely black or completely white. Maybe I could have said that more clearly before, but I’m not particularly worried about clipping highlights/shadows by choosing what are already the brightest/darkest parts of the image.
I normally scan at 1200 dpi. Any temporary files must use a lossless format, e.g. png, tiff (usually), psd, xcf.
You’ll almost always have to use the ‘levels’ controls to adjust the black and white points, as they depend on the paper and ink used. If possible, save the settings as a preset so you can use the same levels on all pages of the booklet so they match. I don’t use the color picker, but just drag the “all channels” black/white point markers to get the brightness correct.
I process the images using blur, smoothing, sharpening, etc. filters - I make heavy use of the G’MIC filters in GIMP - and in the process downscale to 300dpi (¼ the original height/width - this makes CD covers around 1400-1500px depending on crop). Normal screen-printed images don’t have enough detail to make sizes >300dpi useful (although text can be sharper at higher resolution.)
I then save as high quality JPEG (~95%), and upload that to CAA. Since the final image has the screening pattern removed, JPEG doesn’t introduce noticeable quality loss here. (I’m probably gonna start doing PNG instead for future scans.)
Once you get a dark area in a scan to be pure black or a light area to be pure white, other parts of your image are most likely clipping severely. You’re probably not taking it this far, it usually ends up being overly contrasty.
If you’re being more gentle with it you may lose paper texture, noise or detail in some parts of the shadows and highlights, but not everywhere.
In Photoshop’s levels tool you may hold the ALT key pressed while using the sliders to see exactly where the channels are clipping. I found this to be really helpful. Unfortunately Gimp doesn’t have this kind of functionality built in.
Setting shadows and highlights manually so that they’re just about to clip is good practice. You won’t get the blackest blacks and whitest whites this way, but you’re not losing any information either. It keeps the fidelity of your image perfectly intact.
To correct the color manually you could try setting the levels of the individual channels the same way. If the darkest and brightest parts of your image are black and white you should end up with fairly neutral colored shadows and highlights.
If your image is still not contrasty enough at this point, you could use the curves tool to carefully bend the shadows and highlights a little further into saturation without hard clipping anything or messing up the mid tones.
Generally I wouldn’t expect a scan to have perfect blacks and whites. Paper and ink tones are always going to be a little off and uneven. Those are the limitations of the medium. I try to accept them.
I can’t understand this harsh criticism of jpeg. The size reduction is quite significant, even at high quality levels (where I cannot recognize any artifacts). Instead of switching to lossless, better increase the resolution. This gives a much better quality-per-filesize yield.
JPEG was designed for photographs, and does actually still do well for that sort of images. It’s much worse when the image includes sharp edges, like pixel art or anything including text and strongly-defined boxes – like, say, an album cover. It still definitely works, but given that we have an alternative with objectively better quality and not too much larger a file size, the only reason to use JPEG would be if we were vey tight on space. We aren’t. The standard line is “The Cover Art Archive hasn’t complained yet”, and if users care for within their own environment, they’ll be using the smaller image sizes (which offer a bigger savings than the different format) anyway. I can definitely see the data-per-byte argument for personal image libraries, but it’s not really important given how much space we have access to.
Gradients can be quite a problem as well. JPG will turn a smooth gradient into clearly defined bands unless one adds a bit of noise for dithering. That’s why in Photoshop’s gradient tool dither is activated by default.
If there are JPEG artifacts (always talking about high JPEG compression quality), instead of switching to lossless better double the resolution. This will typically result in about the same file size as low resolution lossless, but the information content of the image is much higher (If you don’t believe me: Just try it on you own!)
If there are no visible artifacts, you can take the file size reduction of JPEG for free. This typically occurs when the scanning resolution is high enough, above the original printing resolution (which for archival purposes you should do anyway). The reason is that there are no sharp edges any more and JPEG won’t do any harm.
For cover art scans, this typically means:
If you do 300 DPI lossless, you really should switch to 600 DPI high quality JPEG. About the same file size, but much better images.
If you do 600 DPI lossless or higher, you could just as well go with the file size reduction of JPEG as there won’t be any recognizable artifacts. For the very rare exceptions, 1200 DPI JPEG would again result in much better images than 600 DPI lossless at a similar file size.
For an example, Look at this recent 600 DPI scan @ JPEG qualtiy 90. There are somewhat sharp contrasts at the border and at the logo in the upper right corner. Do you see any artifacts? I do not.
The file size is ~5.5 MB. Lossless PNG would result in ~18 MB, without any recognizable improvement. Lossless PNG at 300 DPI would result in ~5.0 MB, but the original lossy 600 DPI version is much better.
In a nutshell: If there are visible artifacts in high quality JPEG compression, the scanning resolution was not high enough.
Thank you for looking at the image. Usually, I’m setting the black and white point such that the image is just about to clip. However, I don’t care for clipping in areas which are clearly supposed to be all white or all black, as in this case the white parts at the border.
Right, you’re only losing noise and paper texture there. And yes, it was meant to be white, but paper just isn’t perfect. I has a bit of a texture, may be uneven or off-white.
Having paper texture and noise throughout the whole image would be preferable to a tiny bit of extra brightness. Blank spots don’t look good in a noisy environment.
Other than that your scan looks pretty good.
I strongly agree about 600dpi being preferable to 300dpi, though I don’t see the problem with jpg artefacts. Moiré is a much bigger issue.
)
