<?xml version="1.0" encoding="utf-8"?>
<feed xml:lang="en" xmlns="http://www.w3.org/2005/Atom"><title>Recent changes to feature-requests</title><link href="https://sourceforge.net/p/potrace/feature-requests/" rel="alternate"/><link href="https://sourceforge.net/p/potrace/feature-requests/feed.atom" rel="self"/><id>https://sourceforge.net/p/potrace/feature-requests/</id><updated>2022-07-18T20:00:10.975000Z</updated><subtitle>Recent changes to feature-requests</subtitle><entry><title>Minimum segment length</title><link href="https://sourceforge.net/p/potrace/feature-requests/10/" rel="alternate"/><published>2022-07-18T20:00:10.975000Z</published><updated>2022-07-18T20:00:10.975000Z</updated><author><name>Johan Winge</name><uri>https://sourceforge.net/u/alatius/</uri></author><id>https://sourceforge.net1988872089d1099b8ccf6b9c75aa3cd15961f650</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Thank you very much for an excellent tool! I use it to trace very high resolution monochrome bitmaps of font glyphs, where each glyph is constructed by strokes drawn with an elliptical pen. As such, all outlines in the bitmap appear very smooth, and the generated segmentation in Potrace typically contains fairly long segments. However, in those corners where strokes intersect, if the outlines happen to meet in a two pixel wide cusp, a short segment is sometimes created, which ends up in a very small curve, instead of a corner. Please see the attached file for an example. (The actual images I work with have much higher resolution.) This irks me a bit, seeing that this curvature is nothing else than an artefact from how the rasterized contours of the strokes happen to intersect. &lt;br/&gt;
One way that I envision this to be address is to introduce a voluntary step (after vertex adjustments presumably) where a segment shorter than 3 pixels (I think?), that is preceded and followed by longer segments,  will be deleted, and its two vertices are replaced by one placed at the intersection of the extensions of the two surrounding segments (provided that the angle between them is not very small). Do you think this makes sense?&lt;br/&gt;
Would it be feasible to introduce a command line argument to enable something like this?&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Better algorithm for comic-style traces</title><link href="https://sourceforge.net/p/potrace/feature-requests/9/" rel="alternate"/><published>2007-01-20T14:54:22Z</published><updated>2007-01-20T14:54:22Z</updated><author><name>Aaron Optimizer Digulla</name><uri>https://sourceforge.net/u/digulla/</uri></author><id>https://sourceforge.net9fa14f2b87eb0abdabcd7bbd4fa139a107f04d1f</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;The current version of potrace doesn't produce good results with comic style images (i.e. large uniform color areas with some black lines in between).&lt;/p&gt;
&lt;p&gt;Here is an outline of an algorithm for a better trace result:&lt;/p&gt;
&lt;p&gt;- User has to specify the foreground and background color. Usually, this will be black and white.&lt;br /&gt;
- Determine the number of colors in the image and the relative amount of each color.&lt;br /&gt;
- Determine the number of "areas" each color occupies. For an area, there must be more than N pixels of the same color around one spot.&lt;br /&gt;
- Remove all colors from the palette which don't occupy any areas.&lt;br /&gt;
- Copy the image. For all pixels which don't belong to an area (e.g. anti aliasing) fill in the color of the nearest area.&lt;br /&gt;
- Remove the background color from the palette and replace it with a rectangle.&lt;br /&gt;
- Create a trace for the foreground color.&lt;br /&gt;
- Replace any pixel of the foreground color with the color of the closest area.&lt;br /&gt;
- Start with the color which occupies the least amount of total pixels or the least amount of areas in the image and trace it.&lt;br /&gt;
- Replace it with the color of the surrounding areas.&lt;br /&gt;
- Repeat until all colors have been processed.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Stroke line instead of region</title><link href="https://sourceforge.net/p/potrace/feature-requests/8/" rel="alternate"/><published>2006-11-14T11:09:21Z</published><updated>2006-11-14T11:09:21Z</updated><author><name>Romiras</name><uri>https://sourceforge.net/u/romiras/</uri></author><id>https://sourceforge.netdb86c98e9bbe1b8f480069bd6a0ccaea2ef1b9db</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;When i trace some contour line, i get a region, what&lt;br /&gt;
looks like solid line, but it not.&lt;br /&gt;
It would be nice to get thin/wide stroke line instead&lt;br /&gt;
of such region.&lt;/p&gt;
&lt;p&gt;In attached file you can see an image i use for tracing&lt;br /&gt;
contour. You can try to trace it itself do see what&lt;br /&gt;
happens.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Tracing of color images</title><link href="https://sourceforge.net/p/potrace/feature-requests/7/" rel="alternate"/><published>2006-11-10T12:33:56Z</published><updated>2006-11-10T12:33:56Z</updated><author><name>Romiras</name><uri>https://sourceforge.net/u/romiras/</uri></author><id>https://sourceforge.nete99c5d5d1010bc1bf63df949e0128edec9fc91fd</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Add support for color images, at least 8-bit (in&lt;br /&gt;
addition to 1-bit images).&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Reducing nodes</title><link href="https://sourceforge.net/p/potrace/feature-requests/6/" rel="alternate"/><published>2006-11-10T12:26:39Z</published><updated>2006-11-10T12:26:39Z</updated><author><name>Romiras</name><uri>https://sourceforge.net/u/romiras/</uri></author><id>https://sourceforge.net423fe9aca572e88c3850c20492a1e2118253b48e</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Add "Reduce number of nodes" parameter in order to get&lt;br /&gt;
smoothness of result path (like in Corel Trace).&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Centerline switch</title><link href="https://sourceforge.net/p/potrace/feature-requests/5/" rel="alternate"/><published>2006-08-20T16:08:51Z</published><updated>2006-08-20T16:08:51Z</updated><author><name>puddlduk</name><uri>https://sourceforge.net/u/puddlduk/</uri></author><id>https://sourceforge.neteb165fe8a6444c0981b3331db444c33f22585e5f</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;Requesting a centerline option similar to the autotrace&lt;br /&gt;
"--centerline" switch.&lt;/p&gt;
&lt;p&gt;useful for tracing line art, maps/roads, etc.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Font recognition</title><link href="https://sourceforge.net/p/potrace/feature-requests/4/" rel="alternate"/><published>2006-08-08T07:07:55Z</published><updated>2006-08-08T07:07:55Z</updated><author><name>anatoly techtonik</name><uri>https://sourceforge.net/u/techtonik/</uri></author><id>https://sourceforge.neta4077b8768036051a442b127195780227c822482</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;It will be very useful feature to restore text from &lt;br /&gt;
vectorized image into original "letters + font info" &lt;br /&gt;
format.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Add "salt and pepper noise" cleaning to mkbitmap</title><link href="https://sourceforge.net/p/potrace/feature-requests/3/" rel="alternate"/><published>2006-04-20T20:39:26Z</published><updated>2006-04-20T20:39:26Z</updated><author><name>Dov Grobgeld</name><uri>https://sourceforge.net/u/dov-g/</uri></author><id>https://sourceforge.net48d815494bcf0f4471ead4f1d8cf522d43d3d00b</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;When using mkbitmap on a scanned image the resulting&lt;br /&gt;
binary image often contains dust, a.k.a. salt and&lt;br /&gt;
pepper noise. This happens especially with a strong&lt;br /&gt;
high-pass filter. In several applications, e.g. when&lt;br /&gt;
mkbitmap is used on scanned text, there is a natural&lt;br /&gt;
limit to the minimum size of a cluster. It would then&lt;br /&gt;
make sense to erase smaller clusters. There are several&lt;br /&gt;
ways of doing this. E.g. by the opening and closing&lt;br /&gt;
mathematical morphology operators, or by creating a&lt;br /&gt;
label image of the binary image which is then&lt;br /&gt;
histogrammed, and all small clusters are erased.&lt;/p&gt;
&lt;p&gt;(I encountered this problem when working on a scanner&lt;br /&gt;
manager program, with the goal of scanning in gray&lt;br /&gt;
level, and then encoding the resulting pages as djvu&lt;br /&gt;
binary images.)&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Option for totally unsmoothed output</title><link href="https://sourceforge.net/p/potrace/feature-requests/2/" rel="alternate"/><published>2006-02-18T11:41:27Z</published><updated>2006-02-18T11:41:27Z</updated><author><name>Jay Maynard</name><uri>https://sourceforge.net/u/jaymaynard/</uri></author><id>https://sourceforge.net3d143ae343a0291b0aba0691fdb8e434c28c4e05</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;I'm using FontForge to turn a bitmap font into a TrueType font. The font &lt;br /&gt;
is a character set from an old-style IBM CRT. I'd like to retain the &lt;br /&gt;
pixelated look of the font. Specifying -a -1 generates no curves, but still &lt;br /&gt;
smooths away the pixelations. I'd like a way to tell potrace to trace the &lt;br /&gt;
exact outline of the bitmap, jaggies and all. The alternative is manually &lt;br /&gt;
tracing the entire outline of each glyph, which will be a royal pain.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Don't round image boundaries</title><link href="https://sourceforge.net/p/potrace/feature-requests/1/" rel="alternate"/><published>2004-02-27T03:28:34Z</published><updated>2004-02-27T03:28:34Z</updated><author><name>Frederick A Niles</name><uri>https://sourceforge.net/u/gnupooh/</uri></author><id>https://sourceforge.net51968ee68aeebc008eb5d7f2a6daa17330fbfee8</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;I'm using potrace to convert map pieces into eps and&lt;br /&gt;
then put them together.  However, potrace is rounding&lt;br /&gt;
the edge of each piece.  I understand and appreciate&lt;br /&gt;
the fact that potrace rounds edges within the image,&lt;br /&gt;
but I don't think it should round the very edge of the&lt;br /&gt;
original raster image.&lt;/p&gt;
&lt;p&gt;To clarify, if an image has the whole right side black,&lt;br /&gt;
then the output vector eps should have a line going&lt;br /&gt;
from the top to the bottom of the vector image on the&lt;br /&gt;
right side.  Currently the edge will come out slightly&lt;br /&gt;
curved in the vector version, with the corners not painted.&lt;/p&gt;
&lt;p&gt;This is hard to explain, please write back if you don't&lt;br /&gt;
understand.&lt;/p&gt;&lt;/div&gt;</summary></entry></feed>