Interest in scanned decals from kits?

[vsim]

I finally got around to getting a scanner again after not having one for many years. I am planning to scan in the decals for all of the kits I have, mostly because of the realization that by the time I get to some of them the decals may be yellowed or starting to break up. And it would be nice to have backups in case I have issues with putting them on.

So I am curious if there is any general interest in preserving these somewhere (maybe on this site somewhere)? My plan is to calibrate my scanner so that the colors are true/correct in the saved images so that if I go to print them I have the best shot of having the colors correct. I may also look into calibrating my printer as well but that is not as important at this time.

Anyone have any interest in this? Does anyone have a color calibrated printer I could send test images to to see if the colors come out right?

Also, does anyone have any recommendations on how high a resolution to scan? Is 600 ppi sufficient or would it help to scan them in higher than that?

Lastly, I know you can print your own decals at home, but are there any services online that can do single job prints for individuals that are cheap?

Thanks for any suggestions!

[Convectuoso]

Check this thread:

 

 

[Anasazi37]

I use 1200ppi at a minimum for scanning sets. For printing, I use 2400ppi because that's the maximum supported by my printer. I don't bother with magnification because it doesn't really buy you anything, so just 100%. For raster images, you don't generally see a big difference between 1200 and 2400 when printed--especially for older designs. The details just aren't that crisp and the colors can vary a lot. You do see a difference for vectors. I will often use a scanned raster as a base and then redraft the designs as vectors. You get clean lines and solid colors, but it takes time. I do worry about clogging up MW with high res scans, though. The data storage will end up costing Shawn money, potentially a lot.

Single print jobs are generally not cheap and the results will generally not be as good as the original unless someone redrafts the designs. You might be able to do some calibration and color matching between your scanner and an inkjet or laser printer, but it's significantly more difficult (or impossible) to do that for printers that can handle white, like the old ALPS line, which is what you usually need for decals unless you're only doing application on light-colored surfaces. OKI has picked up the white printing baton, but their line is *very* expensive. The cheapest is about $6000 and it doesn't even do a good job. The one that does costs about 5-6 times as much (or so I hear). There are a few folks out there who will do ALPS printing for you, but their reliability is questionable (based on what I've heard from other MW members). I'm retired.

[vsim]

1 hour ago, Convectuoso said:

Check this thread:

 

 

Thanks for the thread link. I saw they had a link to the models section which I had forgotten had low res pictures of decals (maybe some high resolution ones too).

21 minutes ago, Anasazi37 said:

I use 1200ppi at a minimum for scanning sets. For printing, I use 2400ppi because that's the maximum supported by my printer. I don't bother with magnification because it doesn't really buy you anything, so just 100%. For raster images, you don't generally see a big difference between 1200 and 2400 when printed--especially for older designs. The details just aren't that crisp and the colors can vary a lot. You do see a difference for vectors. I will often use a scanned raster as a base and then redraft the designs as vectors. You get clean lines and solid colors, but it takes time. I do worry about clogging up MW with high res scans, though. The data storage will end up costing Shawn money, potentially a lot.

Single print jobs are generally not cheap and the results will generally not be as good as the original unless someone redrafts the designs. You might be able to do some calibration and color matching between your scanner and an inkjet or laser printer, but it's significantly more difficult (or impossible) to do that for printers that can handle white, like the old ALPS line, which is what you usually need for decals unless you're only doing application on light-colored surfaces. OKI has picked up the white printing baton, but their line is *very* expensive. The cheapest is about $6000 and it doesn't even do a good job. The one that does costs about 5-6 times as much (or so I hear). There are a few folks out there who will do ALPS printing for you, but their reliability is questionable (based on what I've heard from other MW members). I'm retired.

Yeah, you are right about the data storage needs for high resolution scans. I don't want to add undue burden onto a hosting site! :). For now I'll just store them locally and see how it works out. Good point about the raster vs. vector scans, I hadn't thought about that. I think my printer maybe only does 600 dpi (color laser). What were you using for printing at 2400 dpi? I'm assuming an inkjet of some form? Or do people still use dye sublimation printers? Right now I'm mostly trying to get the data preserved and am not as worried about needing to replace any decals yet. I just checked my oldest kit which I believe is from the 80s and I was surprised to see the decals were not yellowed or at least it was minimal.

Regardless of hosting, I'd be happy to share any data I gather from my scans. I've got a lot of local storage space. BTW, does anyone know who is maintaining the models section now a days and how to add to it? Chris, you listening?

I've got an IT8 target ordered so I can calibrate my new scanner, and software to do so as well as calibrating a printer. I'll try it out on my color laser just to see how well it works. I have not even remotely looked into specialized decal printers..lol. I was just looking at the micromark decal sheets (clear/white) if I ever needed to print any locally. Any recommendations for doing a good job on the cheap? lol.

[Anasazi37]

On 1/31/2021 at 5:58 PM, vsim said:

Yeah, you are right about the data storage needs for high resolution scans. I don't want to add undue burden onto a hosting site! :). For now I'll just store them locally and see how it works out. Good point about the raster vs. vector scans, I hadn't thought about that. I think my printer maybe only does 600 dpi (color laser). What were you using for printing at 2400 dpi? I'm assuming an inkjet of some form? Or do people still use dye sublimation printers? Right now I'm mostly trying to get the data preserved and am not as worried about needing to replace any decals yet. I just checked my oldest kit which I believe is from the 80s and I was surprised to see the decals were not yellowed or at least it was minimal.

Regardless of hosting, I'd be happy to share any data I gather from my scans. I've got a lot of local storage space. BTW, does anyone know who is maintaining the models section now a days and how to add to it? Chris, you listening?

I've got an IT8 target ordered so I can calibrate my new scanner, and software to do so as well as calibrating a printer. I'll try it out on my color laser just to see how well it works. I have not even remotely looked into specialized decal printers..lol. I was just looking at the micromark decal sheets (clear/white) if I ever needed to print any locally. Any recommendations for doing a good job on the cheap? lol.

I use an ALPS MD-5000P dry resin printer. Hasn't been manufactured in a long time. I'm down to my last one, so I use it sparingly. I actually have to use Windows XP in a virtual machine to run the printer since that was the last OS for which a driver was developed. I also have to use a special cable that converts the old serial printer connector to USB. I'm not entirely sure what kind of ink the newer OKI printers use, but I think it's some form of laser toner. Glad to hear that your old sets are in good shape. Can't say the same for some of mine.

Scanning the sets and archiving them would be a great service for the community. Not sure who is responsible for this section right now.

Let us know how the calibration goes. The micromark paper is probably fine to experiment with--especially at more than $1 per sheet. I have my own special source. Like me, he's basically retired from the business now, but still takes the occasional order. His paper is awesome. Whatever you do, don't use Krylon Crystal Clear to coat what you print, Anyone saying that's a good idea has no clue what they are doing. I've used Microscale Liquid Decal Film for almost 20 years now. If you're okay with translucent designs on clear paper or solid designs on white paper, your laser printer or an inkjet where you let the ink dry thoroughly before coating should be relatively inexpensive.

[pengbuzz]

2 hours ago, Anasazi37 said:

I use an ALPS MD-5000P dry resin printer. Hasn't been manufactured in a long time. I'm down to my last one, so I use it sparingly. I actually have to use Windows XP in a virtual machine to run the printer since that was the last OS for which a driver was developed. I also have to use a special cable that converts the old serial printer connector to USB. I'm not entirely sure what kind of ink the newer OKI printers use, but I think it's some form of laser toner. Glad to hear that your old sets are in good shape. Can't say the same for some of mine.

Scanning the sets and archiving them would be a great service for the community. Not sure who is responsible for this section right now.

Let us know how the calibration goes. The micromark paper is probably fine to experiment with--especially at more than $1 per sheet. I have my own special source. Like me, he's basically retired from the business now, but still takes the occasional order. His paper is awesome. Whatever you do, don't use Krylon Crystal Clear to coat what you print, Anyone saying that's a good idea has no clue what they are doing. I've used Microscale Liquid Decal Film for almost 20 years now. If you're okay with translucent designs on clear paper or solid designs on white paper, your laser printer or an inkjet where you let the ink dry thoroughly before coating should be relatively inexpensive.

Okay...first time I've ever heard not to use Krylon Crystal Clear for coating decals. I'm going to assume based on your comment that it does something rather opposite the intended effect of protecting and sealing it?

Forgive my naivette on this... this particular one is new on me.

[Anasazi37]

6 hours ago, pengbuzz said:

Okay...first time I've ever heard not to use Krylon Crystal Clear for coating decals. I'm going to assume based on your comment that it does something rather opposite the intended effect of protecting and sealing it?

Forgive my naivette on this... this particular one is new on me.

When I started making decals about 20 years ago, I tried Krylon Crystal Clear because that was the advice I was seeing everywhere. I quickly abandoned that approach after a lot of testing. In a nutshell, you're spraying on a solution that after curing will become an inflexible layer on top of another layer that has to be flexible during application, because it's paint that is supposed to be sprayed on a solid surface (e.g. wood, metal). Decal paper doesn't absorb any of the paint like normal paper would, so it just sits on top of the film. As a result, it cracks when the decal paper is flexed. It also tends to be too thick, even if you try multiple light coats. So, you're left with a thicker-than-needed decal that resists application on even the slightest of curved surfaces. That resistance also tends to lead to air bubbles/silvering under the design and makes it difficult to use setting solutions. All of this is worse on small designs since the surface tension is a more significant factor. Last but not least, if you do inkjet printing, you have to be sure that you've covered every last bit of the ink with coating so that it won't dissolve in water, so light coats of Krylon tend not to work well. The only time I've seen the Krylon approach be somewhat okay is when the paint hasn't fully cured and is still a little flexible, but that is a very narrow window of time and it doesn't help with the thickness problem.

I hand-brush LDF on all of the sheets I print to make sure they are fully protected. It dries quickly and results in an extremely thin and flexible layer that adheres to the film. It's also far less wasteful than using a spray can.

[vsim]

On 2/1/2021 at 10:56 PM, Anasazi37 said:

I use an ALPS MD-5000P dry resin printer. Hasn't been manufactured in a long time. I'm down to my last one, so I use it sparingly. I actually have to use Windows XP in a virtual machine to run the printer since that was the last OS for which a driver was developed. I also have to use a special cable that converts the old serial printer connector to USB. I'm not entirely sure what kind of ink the newer OKI printers use, but I think it's some form of laser toner. Glad to hear that your old sets are in good shape. Can't say the same for some of mine.

Scanning the sets and archiving them would be a great service for the community. Not sure who is responsible for this section right now.

Let us know how the calibration goes. The micromark paper is probably fine to experiment with--especially at more than $1 per sheet. I have my own special source. Like me, he's basically retired from the business now, but still takes the occasional order. His paper is awesome. Whatever you do, don't use Krylon Crystal Clear to coat what you print, Anyone saying that's a good idea has no clue what they are doing. I've used Microscale Liquid Decal Film for almost 20 years now. If you're okay with translucent designs on clear paper or solid designs on white paper, your laser printer or an inkjet where you let the ink dry thoroughly before coating should be relatively inexpensive.

Nice, thanks for the pointers.

I'm still waiting on my calibration target to ship, wondering if COVID is causing issues...I hope not, otherwise I'd have to switch to a more expensive source. I haven't had time to mess with it lately anyway due to work, lol.

Anasazi37, you said you'd use 1200 dpi at a minimum. What would you recommend scanning at if I'm going to go through the trouble?

I just looked on micromarks page, and they sell white laser toner now...nice. It's only compatible with 1 specific HP printer but still that's pretty cool. Honestly I'd be fine with just using a sheet of their white decal paper if I needed a white background or paint white under the clear decal.

[Anasazi37]

3 minutes ago, vsim said:

Nice, thanks for the pointers.

I'm still waiting on my calibration target to ship, wondering if COVID is causing issues...I hope not, otherwise I'd have to switch to a more expensive source. I haven't had time to mess with it lately anyway due to work, lol.

Anasazi37, you said you'd use 1200 dpi at a minimum. What would you recommend scanning at if I'm going to go through the trouble?

I just looked on micromarks page, and they sell white laser toner now...nice. It's only compatible with 1 specific HP printer but still that's pretty cool. Honestly I'd be fine with just using a sheet of their white decal paper if I needed a white background or paint white under the clear decal.

2400 ppi, if you have sufficient storage space. It's quadruple the file size compared to 1200 ppi, but we're not talking about huge sheets, or too many of them. Anything higher than that is usually not something a printer can handle, or the human eye can differentiate.

I've seen discussion about the "ghost toner" for that one HP printer for several years now. I think it's made in Germany. One common technique people use is to print out a white version of a design using that toner, then they switch back to the regular toner to print out the color version. My advice would be to do that on two separate sheets, so you end up with a white decal and a color decal. White becomes your background, which you apply first, then put the color one on top of it. Running the white-printed sheet through a second time to get color on top of it generally doesn't work well because the printer can't keep the sheet properly aligned. That also applies to the "cheaper" OKI printer ($6000).

[vsim]

3 minutes ago, Anasazi37 said:

2400 ppi, if you have sufficient storage space. It's quadruple the file size compared to 1200 ppi, but we're not talking about huge sheets, or too many of them. Anything higher than that is usually not something a printer can handle, or the human eye can differentiate.

I've seen discussion about the "ghost toner" for that one HP printer for several years now. I think it's made in Germany. One common technique people use is to print out a white version of a design using that toner, then they switch back to the regular toner to print out the color version. My advice would be to do that on two separate sheets, so you end up with a white decal and a color decal. White becomes your background, which you apply first, then put the color one on top of it. Running the white-printed sheet through a second time to get color on top of it generally doesn't work well because the printer can't keep the sheet properly aligned. That also applies to the "cheaper" OKI printer ($6000).

Gotcha...yeah I should be able to do 2400, it'll suck up some space but shouldn't be too big a deal. It will just take me a while to scan them all, lol. What can I say, I'm a data hoarder :D.

[vsim]

Finally got my IT8 target and got everything calibrated after fighting scanner software for a while. I've now learned more about color calibration than I ever really wanted to know, lol. But I've got 3 different scanning programs producing the same results now. I had to make a custom calibration for each software program because apparently they all have different definitions for what RAW output is, lol. That took me a while to figure out. The fun part is, only 1 of the programs works reliably at 2400 dpi, but it doesn't even do lossless compression on the output tif file so I may have to post process with Gimp or something just to keep file sizes down a bit.

So now I need to find time to scan a bunch of kits, lol...oy...

[Anasazi37]

1 hour ago, vsim said:

Finally got my IT8 target and got everything calibrated after fighting scanner software for a while. I've now learned more about color calibration than I ever really wanted to know, lol. But I've got 3 different scanning programs producing the same results now. I had to make a custom calibration for each software program because apparently they all have different definitions for what RAW output is, lol. That took me a while to figure out. The fun part is, only 1 of the programs works reliably at 2400 dpi, but it doesn't even do lossless compression on the output tif file so I may have to post process with Gimp or something just to keep file sizes down a bit.

So now I need to find time to scan a bunch of kits, lol...oy...

I've been building my own 3D laser triangulation scanner for the past several months and it's been a similar experience: learned more about how all of that stuff works than I ever really wanted to know. Sounds like you've made significant progress. Keep us posted!

[vsim]

On 3/21/2021 at 5:37 PM, Anasazi37 said:

I've been building my own 3D laser triangulation scanner for the past several months and it's been a similar experience: learned more about how all of that stuff works than I ever really wanted to know. Sounds like you've made significant progress. Keep us posted!

Slowly but surely. And now I've got an automated set of routines that can patch together box photos since they are too big for my scanner. It works surprisingly well even automated. (The program is called hugin). Still haven't completely automated that process but I am fairly close.

@Anasazi37, I've noticed my my current workflows that I'm making that when I use hugin to patch together pictures, it converts it from a 24-bit TIFF to a 32-bit TIFF. I think what's happening is it's going from "TrueColor" to "TrueColorAlpha" or at least that's what Image Magick says, and I have no idea how to tell it to keep it at 24 bpp. I presume it is adding more data to an alpha channel. Do you think I should just go ahead and convert my whole process over to use 32-bit or stick with 24-bit knowing that the box scans are going to get converted to 32-bit anyway. I'm not sure how much bigger then 2400 dpi decal scans would get filesize-wise, maybe 25% bigger? I'm just not sure if the extra bits are worth it or if it's beyond what anyone would ever notice anyway.

The 2400 dpi Sv-262Hs Hasegawa decals are already 418MB with LZW compression, lol.

If anyone has a specific kit they'd like to see data for let me know. I have no particular order for scanning these and it will take me a while!

[Anasazi37]

21 hours ago, vsim said:

Slowly but surely. And now I've got an automated set of routines that can patch together box photos since they are too big for my scanner. It works surprisingly well even automated. (The program is called hugin). Still haven't completely automated that process but I am fairly close.

@Anasazi37, I've noticed my my current workflows that I'm making that when I use hugin to patch together pictures, it converts it from a 24-bit TIFF to a 32-bit TIFF. I think what's happening is it's going from "TrueColor" to "TrueColorAlpha" or at least that's what Image Magick says, and I have no idea how to tell it to keep it at 24 bpp. I presume it is adding more data to an alpha channel. Do you think I should just go ahead and convert my whole process over to use 32-bit or stick with 24-bit knowing that the box scans are going to get converted to 32-bit anyway. I'm not sure how much bigger then 2400 dpi decal scans would get filesize-wise, maybe 25% bigger? I'm just not sure if the extra bits are worth it or if it's beyond what anyone would ever notice anyway.

The 2400 dpi Sv-262Hs Hasegawa decals are already 418MB with LZW compression, lol.

If anyone has a specific kit they'd like to see data for let me know. I have no particular order for scanning these and it will take me a while!

You're using hugin, eh? It's one of many photogrammetry-inspired applications out there that can stitch (mosaic) images together. They all pretty much do the same thing. Their automated workflows tend to a do a decent job, but not always a perfect one, so long as there is decent overlap between the images, the images were collected under similar conditions, and the images contain enough "interesting" stuff for the underlying computer vision routines to find and link across them. This kind of image processing is a subset of what I do for a living, so I can geek out over it in a big way if I'm not careful.

I think you're right about hugin using the alpha channel (RGBA). My best guess as to why it's doing so is that it needs to know which pixels contain useful information and which pixels are just background fill that can be ignored. This is important during mosaicking because the software might have to subtly warp the images to make them fit together well and when it does that, background fill is introduced. If you don't ignore that stuff, you end up with lots of useless and annoying black areas in the end result instead of something seamless. The alpha channel is where the software keeps track of that, where a pixel value of 0 means "ignore" and a value of 255 means "keep." The software, after mosaicking, likely just spits out the RGBA version because the end result is going to have some background fill around the edges. My advice is to let hugin do its thing, then use another application to remove the alpha channel (convert from RGBA to RGB). You'll have larger files in the interim, but once the RGB version is created, you can delete the interim stuff.

[vsim]

2 hours ago, Anasazi37 said:

You're using hugin, eh? It's one of many photogrammetry-inspired applications out there that can stitch (mosaic) images together. They all pretty much do the same thing. Their automated workflows tend to a do a decent job, but not always a perfect one, so long as there is decent overlap between the images, the images were collected under similar conditions, and the images contain enough "interesting" stuff for the underlying computer vision routines to find and link across them. This kind of image processing is a subset of what I do for a living, so I can geek out over it in a big way if I'm not careful.

I think you're right about hugin using the alpha channel (RGBA). My best guess as to why it's doing so is that it needs to know which pixels contain useful information and which pixels are just background fill that can be ignored. This is important during mosaicking because the software might have to subtly warp the images to make them fit together well and when it does that, background fill is introduced. If you don't ignore that stuff, you end up with lots of useless and annoying black areas in the end result instead of something seamless. The alpha channel is where the software keeps track of that, where a pixel value of 0 means "ignore" and a value of 255 means "keep." The software, after mosaicking, likely just spits out the RGBA version because the end result is going to have some background fill around the edges. My advice is to let hugin do its thing, then use another application to remove the alpha channel (convert from RGBA to RGB). You'll have larger files in the interim, but once the RGB version is created, you can delete the interim stuff.

Yeah, that makes sense. Thanks! I figured out how to get Image Magick to remove the alpha channel and get rid of the extraneous virtual canvas that hugin had created as well. So that's all automated now as well. Thanks for the info, it was just enough so that I could squeak by with google searches and experimenting with different options. Half the time I don't know the terminology used for graphics programs all that well so I'm not always certain what to search for to solve my issue.

What you said about the alpha channel makes sense, thanks for explaining, I was wondering why it had 1 bit for the alpha channel, lol.

So my scanning software goes straight from 24-bit to 48-bit for colors so I think I will keep it at 24-bit color scans. And so far I have been disabling any descreening or anything else like that in the scans. I figure that can always be done postprocess, but if I lose the data during the original scans I can't get it back. I'm mostly scanning the boxes just because they have nice art. The decals are more to try and preserve.

Now I just need to automate things a little bit more so I don't have to edit scripts every time I run. And then see if the final scans from hugin are the approximate correct physical dimensions of the original boxes, etc. as it seems to mess that up some. I got it set for the test images I did but I'm not sure it will be correct for other images. It's not super important as it will only affect box art scans, but I'd still like it to match if I can get it to.

And then I need to get back to building my VF-0B kit! lol

[Photogirl]

I'm looking for an actual copy of this sheet.
I know i can scan it and print it out on printer slide paper, but if i can find the original, id prefer it. 

[vsim]

18 hours ago, Photogirl said:

I'm looking for an actual copy of this sheet.
I know i can scan it and print it out on printer slide paper, but if i can find the original, id prefer it. 

That looks like it's from the Hasegawa kit 65763 Macross VF-1J Max & Miria. I have it and could scan it for you but it doesn't sound like you want a scan.

[vsim]

Did some more testing and tweaking this morning, scanned in the VF-0B kit I'm mid progress on. Unfortunately it means I've already but some of the decals out..oh well 

VF-0B Box Cover stitched together with Hugin (and shrunk down for posting).

 

Edited March 29, 2021 by vsim
uploading smaller image

[Bolt]

Nice job looks good. 

[TMBounty_Hunter]

You know, stitching scans together manually really isn't that hard or time consuming. I've spent my teen years doing that with a lot of my kit boxes as well as magazines. Newtype being too wide for a standard scanner was infuriating. But after a bit of practice whether it's x4 scans for large box, or x8 scans for a REALLY large box like a PG, it's really not too bad. Now I just have an A3 scanned and life is twice as good. Maybe I'm just prejudiced against automation because every time I've tried it the results needed more fixing than just starting manually from scratch. That plus when working manually I know that the mistakes would be my own and it motives me to avoid them since once you put it on the internet it's there forever and every time someone reposts the image that mistake will scream back at you. Plus you never know how those built up photoshop skills might be useful for other things.

Here's 3 samples of large boxes that all needed x4 scans on my old HP scanner that maxed out at 600dpi. About 10-25min of manual work each, filtered and 25% resolution for easy upload.

 

This PG Eva box was x8 scans and maybe an hour of work but most of that was dust and scratch removal because this box was far from new when I got it. Again 600dpi, filtered and 33% res to keep size down.

The most important part is getting good raw scans without distortion and with plenty of overlap because regardless if you're going it manually or using a software black box GIGO law is absolute. When scanning Hasegawa boxes too long for a standard scanner I highly recommend a stack of the old format Hobby Japan magazines, they are the absolutely perfect size to put inside the box to flatten it against the scanner glass and minimize distortion from one edge being elevated off the edge of the bed.

[vsim]

On 3/29/2021 at 12:38 PM, TMBounty_Hunter said:

You know, stitching scans together manually really isn't that hard or time consuming. I've spent my teen years doing that with a lot of my kit boxes as well as magazines. Newtype being too wide for a standard scanner was infuriating. But after a bit of practice whether it's x4 scans for large box, or x8 scans for a REALLY large box like a PG, it's really not too bad. Now I just have an A3 scanned and life is twice as good. Maybe I'm just prejudiced against automation because every time I've tried it the results needed more fixing than just starting manually from scratch. That plus when working manually I know that the mistakes would be my own and it motives me to avoid them since once you put it on the internet it's there forever and every time someone reposts the image that mistake will scream back at you. Plus you never know how those built up photoshop skills might be useful for other things.

Here's 3 samples of large boxes that all needed x4 scans on my old HP scanner that maxed out at 600dpi. About 10-25min of manual work each, filtered and 25% resolution for easy upload.

 

This PG Eva box was x8 scans and maybe an hour of work but most of that was dust and scratch removal because this box was far from new when I got it. Again 600dpi, filtered and 33% res to keep size down.

The most important part is getting good raw scans without distortion and with plenty of overlap because regardless if you're going it manually or using a software black box GIGO law is absolute. When scanning Hasegawa boxes too long for a standard scanner I highly recommend a stack of the old format Hobby Japan magazines, they are the absolutely perfect size to put inside the box to flatten it against the scanner glass and minimize distortion from one edge being elevated off the edge of the bed.

I've got a 1 year old in the house, I don't usually get time to stitch images together, lol.

I think I discovered what you just said above. I've started putting manga, etc. inside the box to keep it flat on the scanner. That seems to help the images stitch together automatically better. Although I still had 1 or 2 where I needed to go in and manually futz with it some. Hugin seems to not be happy sometimes if it needs to stretch the image, so I think I need to put more effort into making sure the box is completely flat and not scan near the edge where it is riding up on the lip of the scanner. Sometimes Hugin does great with the automation I've set up, other times it does weird things....still figuring the tool out and all the command line arguments; but slowly getting better at figuring out exactly what it's doing. If I ever get it where it works very reliably I'd happily share my scripts.

I'm only doing the boxes at 400 dpi for now as it seems to give a nice image without taking up nearly as much space as the 2400 dpi decals, lol.

I discovered an issue with the VF-0B scans, if you look, you can see in the "Macross Zero" lettering in the older image above where it didn't stitch together correctly, although I don't know why.

Here's a better one

I've also fixed my scans of the side of the box which had similar issues. I'm not sure if too much overlap is causing the issues, or what exactly Hugin's issue is. I've started scanning black space around the box to try to give it a better defined border to adjust to, but that may make things worse.

Here's the Sv-262 that I fixed today as well:

@TMBounty_Hunter, what software did you use to stitch yours together? Gimp or something similar? 

[Anasazi37]

On 3/28/2021 at 1:23 PM, vsim said:

Yeah, that makes sense. Thanks! I figured out how to get Image Magick to remove the alpha channel and get rid of the extraneous virtual canvas that hugin had created as well. So that's all automated now as well. Thanks for the info, it was just enough so that I could squeak by with google searches and experimenting with different options. Half the time I don't know the terminology used for graphics programs all that well so I'm not always certain what to search for to solve my issue.

What you said about the alpha channel makes sense, thanks for explaining, I was wondering why it had 1 bit for the alpha channel, lol.

So my scanning software goes straight from 24-bit to 48-bit for colors so I think I will keep it at 24-bit color scans. And so far I have been disabling any descreening or anything else like that in the scans. I figure that can always be done postprocess, but if I lose the data during the original scans I can't get it back. I'm mostly scanning the boxes just because they have nice art. The decals are more to try and preserve.

Now I just need to automate things a little bit more so I don't have to edit scripts every time I run. And then see if the final scans from hugin are the approximate correct physical dimensions of the original boxes, etc. as it seems to mess that up some. I got it set for the test images I did but I'm not sure it will be correct for other images. It's not super important as it will only affect box art scans, but I'd still like it to match if I can get it to.

And then I need to get back to building my VF-0B kit! lol

Oh good, I was thinking that it would be pretty easy to remove the alpha channel in ImageMagick, but hadn't looked into it. Graphic design, computer vision, photogrammetry, and remote sensing people all use different terms for pretty much the same stuff. You basically have to speak four different technical languages to find useful information online. Only 1 bit for the alpha channel? Interesting. Really must have been a binary mask (1 = use, 0 = don't use). You don't see that very often. Usually it's a gradient value between 0 and 1 or 0 and 255. I don't see a need to go beyond 24-bit for this kind of project. 16 bits per color channel is excessive and printers aren't really set up to deal with it. Keeping more raw scans that you can then postprocess in different ways is a good idea. I love that you're building enough flexibility into your scripts that you won't have to make changes each time you run. Command line options FTW. Bash, PowerShell, or something else?

On 3/29/2021 at 10:38 AM, TMBounty_Hunter said:

You know, stitching scans together manually really isn't that hard or time consuming. I've spent my teen years doing that with a lot of my kit boxes as well as magazines. Newtype being too wide for a standard scanner was infuriating. But after a bit of practice whether it's x4 scans for large box, or x8 scans for a REALLY large box like a PG, it's really not too bad. Now I just have an A3 scanned and life is twice as good. Maybe I'm just prejudiced against automation because every time I've tried it the results needed more fixing than just starting manually from scratch. That plus when working manually I know that the mistakes would be my own and it motives me to avoid them since once you put it on the internet it's there forever and every time someone reposts the image that mistake will scream back at you. Plus you never know how those built up photoshop skills might be useful for other things.

The most important part is getting good raw scans without distortion and with plenty of overlap because regardless if you're going it manually or using a software black box GIGO law is absolute. When scanning Hasegawa boxes too long for a standard scanner I highly recommend a stack of the old format Hobby Japan magazines, they are the absolutely perfect size to put inside the box to flatten it against the scanner glass and minimize distortion from one edge being elevated off the edge of the bed.

Manual works if your scans are completely distortion-free, as you point out. I've been doing this kind of processing on imagery for about 15 years now, professionally, and what most of us in that world end up using is a combination of automated and manual. Automated quickly takes care of about 90-95% of the work, then you can manually tweak a few spots to fix small misalignment issues. But even the automated approach will benefit from distortion-free scans. Means less work overall. If an image is shifted or rotated, that's an easy fix either way. If there are deformities of different sizes and types because the original image isn't perfectly flat, that's a lot harder for a person, or even software, to mitigate.

Edited April 3, 2021 by Anasazi37

[TMBounty_Hunter]

On 4/3/2021 at 3:28 PM, vsim said:

@TMBounty_Hunter, what software did you use to stitch yours together? Gimp or something similar? 

Photoshop CS4 cuz that's what I had for the longest time and Adobe and kinda shits these days with their practices

 

  

On 4/3/2021 at 5:32 PM, Anasazi37 said:

Manual works if your scans are completely distortion-free, as you point out. 

Hmm I don't know how that came across but I meant the opposite. I've found manual work to be pretty much the only way to fix distortions. Automations seem to just amplify them resulting in more work than starting manually from scratch.

Maybe this has to do with resolution. It's very easy to use the printing pattern of the image to align things manually at 600dpi but maybe to machines that's just noise.

Edited April 7, 2021 by TMBounty_Hunter

[Anasazi37]

13 hours ago, TMBounty_Hunter said:

Hmm I don't know how that came across but I meant the opposite. I've found manual work to be pretty much the only way to fix distortions. Automations seem to just amplify them resulting in more work than starting manually from scratch.

Maybe this has to do with resolution. It's very easy to use the printing pattern of the image to align things manually at 600dpi but maybe to machines that's just noise.

Depends on the nature of the distortion. The options available in Photoshop tend to be pretty limited, what I'll call "affine" because that is fairly common way to describe them. Essentially combinations of rotation, scale, and translation. That includes grabbing corners of an image and "stretching" them out to improve alignment. For flatbed scans, that's generally sufficient because of how the optics of the system work. A robust automated approach can nail that kind of task, but off-the-shelf graphics software generally doesn't excel in this area. The situation becomes way more complicated if you take actual pictures with a camera and want to stitch those together seamlessly. Goes from a two-dimensional problem to a three-dimensional one.

[TMBounty_Hunter]

8 hours ago, Anasazi37 said:

The situation becomes way more complicated if you take actual pictures with a camera and want to stitch those together seamlessly. Goes from a two-dimensional problem to a three-dimensional one.

Honestly that just seems like an exercise in self flagellation

[Anasazi37]

10 hours ago, TMBounty_Hunter said:

Honestly that just seems like an exercise in self flagellation

There are times when something like that makes sense, but not for decals or box art. Why punch yourself in the head when you don't have to?

[vsim]

On 4/3/2021 at 5:32 PM, Anasazi37 said:

Oh good, I was thinking that it would be pretty easy to remove the alpha channel in ImageMagick, but hadn't looked into it. Graphic design, computer vision, photogrammetry, and remote sensing people all use different terms for pretty much the same stuff. You basically have to speak four different technical languages to find useful information online. Only 1 bit for the alpha channel? Interesting. Really must have been a binary mask (1 = use, 0 = don't use). You don't see that very often. Usually it's a gradient value between 0 and 1 or 0 and 255. I don't see a need to go beyond 24-bit for this kind of project. 16 bits per color channel is excessive and printers aren't really set up to deal with it. Keeping more raw scans that you can then postprocess in different ways is a good idea. I love that you're building enough flexibility into your scripts that you won't have to make changes each time you run. Command line options FTW. Bash, PowerShell, or something else?

Working in Windows, I don't have time to setup a Linux box at home, although I've been using Linux since 1993 so I'm one of the old timers on it, lol. So no bash for me unless I setup Mingw or something similar. Batch and Powershell seem to be working well for me so far. The merging script it just a batch script that I edit the filenames in each time, could easily be converted to take them as command line args. My powershell script is one I use to convert all tifs in a directory to have LZW compression so that I can get the decal scans down in side and not have 0.5 - 1 Gig per scan, lol.

[Anasazi37]

On 4/11/2021 at 8:27 AM, vsim said:

Working in Windows, I don't have time to setup a Linux box at home, although I've been using Linux since 1993 so I'm one of the old timers on it, lol. So no bash for me unless I setup Mingw or something similar. Batch and Powershell seem to be working well for me so far. The merging script it just a batch script that I edit the filenames in each time, could easily be converted to take them as command line args. My powershell script is one I use to convert all tifs in a directory to have LZW compression so that I can get the decal scans down in side and not have 0.5 - 1 Gig per scan, lol.

Nicely done