Talk:Image tracing

Broken Link

I don't know how to fix it, so I'm passing this on. The link under the middle image, "The photograph at left vectorized with RaveGrid" leads to http://upload.wikimedia.org/wikipedia/en/f/f5/SilversmithRaveGrid.svg and it's broken. Leedeth 07:25, 12 November 2006 (UTC)eid k baad[reply]

Done EoRdE6 (talk) 19:19, 9 October 2014 (UTC)[reply]

Example

The image here is a strange example of vectorization. As noted, file sizes of vector versions are much larger, and the image is distorted, but of course, it is then scalable. Simple graphics tend to lend themselves better to vectorization. Perhaps I (or someone else) should add another example of vectorization to demonstrate it's use for simpler images, like a hand-inked logo. -Ames (talk) 01:26, 7 February 2009 (UTC)[reply]

I second this notion. Perhaps Erin Silversmith and Shaddim (and whomever I missed) still have access to the software they used to create these images. I think a rasterized version of some free black-and-white logo would also do, say the GNU logo. --Mudd1 (talk) 19:06, 23 February 2014 (UTC)[reply]

What about signatures as example for "useful" (and boring easy) vectorization? The Silversmith vectorization exmaple was chosen to illustrate limits and variations of the vectorization approaches. In a signature example most vectorization software will behave pretty well (and are therefore indistinguishable).

Schriftzug des Christoph Columbus als JPEG-Rastergrafik (1,308 × 481 Pixel), 63 kB
Vektorisierte, zweifarbige Variante, 19 kB

see also the deWP page. Shaddim (talk) 14:41, 24 February 2014 (UTC)[reply]

Yeah, good point and good example. --Mudd1 (talk) 14:54, 27 February 2014 (UTC)[reply]

Instructions

Could we potentially use this article to provide instructions for users wanting to convert Wikipedia's database of raster images into vector ones? We would need to add a link to the template but first I think this article needs to be much improved.Spitfire19 (Talk) 13:54, 3 May 2010 (UTC)[reply]

Merge or change disambiguation

After discussion at Talk:Vectorization (parallel computing), I propose this page get renamed to "Vectorization (vector graphics)" or even merge this article in to vector graphics. The reason being the broad subject of "computer graphics" can include both kinds of vectorization (e.g., vectorized instructions within a GPU). If no one objects, I'll rename this page. —Ben FrantzDale (talk) 13:46, 8 November 2010 (UTC)[reply]

Isn't "Vectorization (vector graphics)" rather redundantly redundant. How about just "Vector tracing"? AnonMoos (talk) 16:33, 8 November 2010 (UTC)[reply]

It is redundant in a way, but I like "vectorization" as the primary title and the best disambiguator I can think of is "the kind of vectorization involved in vector graphics". I considered "Vectorization (digital illustration)" and am open to others. "Bector tracing" is OK, I guess, but it seems like a less-common word for the process. —Ben FrantzDale (talk) 17:40, 8 November 2010 (UTC)[reply]

Then "Vectorization (computer image tracing)" would avoid falling under the department of redundancy department. AnonMoos (talk) 13:12, 9 November 2010 (UTC)[reply]

Sure. It is a little unusual for a disambiguation in that it refines "vectorization" more than it gives the field in which we are talking about "vectorization", but it can be read either way and I agree about the DoRD issue. I'll rename it. —Ben FrantzDale (talk) 16:57, 9 November 2010 (UTC)[reply]

I'll actually do Vectorization (image tracing) since it could be done manually with no computer (e.g., the initial Utah teapot). —Ben FrantzDale (talk) 16:58, 9 November 2010 (UTC)[reply]

List of known algorithms

This article somewhat desperately needs a list of vectorization algorithms, so that people can find concrete examples of what was discussed. It's a fairly low-information article as it stands, and doesn't really tell you much about what the general approaches to vectorization are, which algorithms exist, which algorithms out of that pool are actually used a lot, and perhaps some information that links vectorization to related subject fields like edge detection/image kerning, non-linear classifiers, etc - Pomax, 9 October 2014 — Preceding unsigned comment added by 75.156.54.244 (talk) 16:52, 9 October 2014 (UTC)[reply]

Had to be re-added after blanking, I don't have specific rollback features EoRdE6 (talk) 19:12, 9 October 2014 (UTC)[reply]

How-to instructions and original research

Although I really like this article – because it's useful! – that may (bizarrely) conflict with a WP policy on "Wikipedia is not a user guide" (or "how-to manual") or some such.

Also, it seems to me that the unattributed examples result from the individual efforts of several helpful editors, rather than being "drawn" (heh) from secondary sources. While, of course, editors must provide some creative input if they are to write a good explanation of a topic, I wonder whether these examples might go a bit too far in the direction of original research? I hesitate to put it more strongly than that, because I love to encourage all good contributions that improve Wikipedia. Perhaps there exist some reliable sources that come to similar conclusions about the range of useful applicability, and limitations of, the various programs mentioned? yoyo (talk) 00:14, 26 February 2017 (UTC)[reply]

Whiel I see your point, I think this is still inside the freedom we have as WP authors. Especially as this is not an controversial topic or a article about a person, lower standards for refs apply here. Also, this is not really "Original (new) research" as it reflects just state of the art understanding and technology. Still, it would be preferable if more sources could be found and added. Shaddim (talk) 19:25, 27 February 2017 (UTC)[reply]

I don't see the article as being a user guide or a how-to manual. It is not telling people how they should use a particular software application. The article is making statements about how such software is used in practice because image tracing is not a magic bullet. The software has limitations, and the way users get around those limitations includes manually tracing an image and then using software to trace that binary image. If image tracing applications produced good results, a bot would convert a lot of images on Commons to SVG.

While there is a prohibition on original research, WP editors are empowered to do WP:CALCulations and provide examples. While I think the vectorization of the continuous tone images is a bit silly, it does show that the traced results are poor and that all the traced image sizes pale in comparison to the 25 kB original JPEG. The Radula trace is more practical, but it shows the results are not ideal; outlines are lost.

Yes, I'd like better sources, but I don't see the article straying into forbidden territory. Glrx (talk) 21:34, 27 February 2017 (UTC)[reply]

Example images and outdated filesize references

Some of the images used as examples have apparently been edited following their being referenced in this article. So far these are:

The article references filesizes, which are subject to change if images are re-encoded. JPG, PNG and SVG files are all potentially subject to such re-encoding, for example by lossless image optimizers like jpegtran or pngcrush.

Either we'll need a consensus that filesizes are "per original encoding output by the tracing software", rely on someone occasionally checking and updating filesize references whenever the referenced files change, or have dedicated reference files tagged with Template:Original to warn editors not to perform edits to the files. djr13 (talk) 22:19, 22 April 2020 (UTC)[reply]

I think it would be better to have dedicated images. That way we could compare the result of applying tracing program X with the result of applying tracing program Y. AnonMoos (talk) 23:19, 24 April 2020 (UTC)[reply]