Loading ...
Sorry, an error occurred while loading the content.

Re: [AutoTrace] Re: autotrace 0.23 (0.24 problem with closed areas).

Expand Messages
  • Ian MacPhedran
    ... If that refers to my erlier message, I was referring to _output_ formats, not input formats. With more words: It appears to be the postscript output driver
    Message 1 of 20 , Sep 21, 2000
    • 0 Attachment
      On Thu, 21 Sep 2000 Ted.Harding@... wrote:

      > The report that it depends on the image source format is puzzling ...

      If that refers to my erlier message, I was referring to _output_ formats,
      not input formats. With more words:

      It appears to be the postscript output driver that is having the problem.
      Output created using the fig or p2e output drivers seems okay.

      Ian.
      ----------------------------------------------------------------------------
      Ian MacPhedran, Engineering Computer Centre, 2B13 Engineering Building,
      University of Saskatchewan, 57 Campus Drive, Saskatoon SK S7N 5A9, CANADA
      Phone: (306)966-4832 Fax: (306)966-5205 Email: Ian_MacPhedran@...
    • Stewart C. Russell
      ... Hmm ;-( EPS is the only output format I ve used, and indeed have any use for. I m glad I ve still got an 0.21 binary on my home system. Stewart
      Message 2 of 20 , Sep 22, 2000
      • 0 Attachment
        Ian MacPhedran wrote:
        >
        > Seems to be driver specific - p2e and fig seem to get it right.

        Hmm ;-(

        EPS is the only output format I've used, and indeed have any use for.
        I'm glad I've still got an 0.21 binary on my home system.

        Stewart
      • Bernhard Herzog
        ... AFAICT, this has something to do with the new support for color images. Earlier versions only support B&W images and constructed a list of paths that were
        Message 3 of 20 , Sep 22, 2000
        • 0 Attachment
          "Stewart C. Russell" <stewart@...> writes:

          > 0.24 seems to
          > get the inside of closed areas wrong, even on PBM input. I don't think
          > this is an architecture thing -- it happens on both x86-linux and
          > Sparc-Solaris.

          AFAICT, this has something to do with the new support for color images.

          Earlier versions only support B&W images and constructed a list of paths
          that were to be filled as a whole. In PostScript terms each of the list
          items consists of one moveto followed only by lineto's and curveto's adn
          all list items together form a postscript path that is filled.

          Autotrace now supports color input and each of the splines has a color
          parameter. When I updated the EPS, SVG and SK export code, I assumed
          that each of the paths would be filled individually and it worked for
          the images I tested it with. So I could be considered responsible for the
          Bugs.

          The best solution would be for autotrace to mark the splines that should
          be filled together. As a workaround it might be possible to just treat
          consecutive splines with the same fill color as subpath of the same
          postscript path and only start a new postscript path if the color
          changes. For SVG and SK simillar solutions should work.

          I've included a patch for the EPS code below.

          Bernhard

          PS: Martin, is there a list of contributers somewhere? There seem to
          have been several contributers by now, especially for output filters.
          The names I found by looking at a few files were in no particular order
          and probably incomplete:

          Kevin O' Gorman
          Reini Urban
          Ian MacPhedran
          Wolfgang Glunz
          Masatake YAMATO
          Bernhard Herzog


          Here the patch:

          *** output-eps.c.~1~ Tue Sep 19 20:19:04 2000
          --- output-eps.c Fri Sep 22 16:10:06 2000
          ***************
          *** 122,127 ****
          --- 122,128 ----
          out_splines (FILE * ps_file, spline_list_array_type shape)
          {
          unsigned this_list;
          + color_type last_color;

          for (this_list = 0; this_list < SPLINE_LIST_ARRAY_LENGTH (shape);
          this_list++)
          ***************
          *** 132,150 ****
          spline_list_type list = SPLINE_LIST_ARRAY_ELT (shape, this_list);
          spline_type first = SPLINE_LIST_ELT (list, 0);

          ! c = k = 255 - list.color.r;
          ! m = 255 - list.color.g;
          ! if (m < k)
          ! k = m;
          ! y = 255 - list.color.b;
          ! if (y < k)
          ! k = y;
          ! c -= k;
          ! m -= k;
          ! y -= k;
          ! /* symbol k is used for CorelDraw 3/4 compatibility */
          ! OUT4 ("%f %f %f %f k\n", (double) c/255.0,
          ! (double) m/255.0,(double) y/255.0, (double) k/255.0);
          OUT_COMMAND2 (START_POINT (first).x, START_POINT (first).y, "m");

          for (this_spline = 0; this_spline < SPLINE_LIST_LENGTH (list);
          --- 133,156 ----
          spline_list_type list = SPLINE_LIST_ARRAY_ELT (shape, this_list);
          spline_type first = SPLINE_LIST_ELT (list, 0);

          ! if (this_list == 0 || !COLOR_EQUAL(list.color, last_color))
          ! {
          ! if (this_list > 0)
          ! OUT_LINE ("f");
          ! c = k = 255 - list.color.r;
          ! m = 255 - list.color.g;
          ! if (m < k)
          ! k = m;
          ! y = 255 - list.color.b;
          ! if (y < k)
          ! k = y;
          ! c -= k;
          ! m -= k;
          ! y -= k;
          ! /* symbol k is used for CorelDraw 3/4 compatibility */
          ! OUT4 ("%f %f %f %f k\n", (double) c/255.0,
          ! (double) m/255.0,(double) y/255.0, (double) k/255.0);
          ! }
          OUT_COMMAND2 (START_POINT (first).x, START_POINT (first).y, "m");

          for (this_spline = 0; this_spline < SPLINE_LIST_LENGTH (list);
          ***************
          *** 160,167 ****
          END_POINT (s).x, END_POINT (s).y,
          "c");
          }
          ! OUT_LINE ("f");
          }
          }


          --- 166,175 ----
          END_POINT (s).x, END_POINT (s).y,
          "c");
          }
          ! last_color = list.color;
          }
          + if (SPLINE_LIST_ARRAY_LENGTH (shape) > 0)
          + OUT_LINE ("f");
          }





          --
          Bernhard Herzog | Sketch, a drawing program for Unix
          herzog@... | http://sketch.sourceforge.net/
        • Martin Weber
          Who is missing on the contributors list?
          Message 4 of 20 , Sep 22, 2000
          • 0 Attachment
            Who is missing on the contributors list?

            Bernhard Herzog schrieb:

            >
            > "Stewart C. Russell" <stewart@...> writes:
            >
            > > 0.24 seems to
            > > get the inside of closed areas wrong, even on PBM input. I don't think
            > > this is an architecture thing -- it happens on both x86-linux and
            > > Sparc-Solaris.
            >
            > AFAICT, this has something to do with the new support for color images.
            >
            > Earlier versions only support B&W images and constructed a list of paths
            > that were to be filled as a whole. In PostScript terms each of the list
            > items consists of one moveto followed only by lineto's and curveto's adn
            > all list items together form a postscript path that is filled.
            >
            > Autotrace now supports color input and each of the splines has a color
            > parameter. When I updated the EPS, SVG and SK export code, I assumed
            > that each of the paths would be filled individually and it worked for
            > the images I tested it with. So I could be considered responsible for the
            > Bugs.
            >
            > The best solution would be for autotrace to mark the splines that should
            > be filled together. As a workaround it might be possible to just treat
            > consecutive splines with the same fill color as subpath of the same
            > postscript path and only start a new postscript path if the color
            > changes. For SVG and SK simillar solutions should work.
            >
            > I've included a patch for the EPS code below.
            >
            > Bernhard
            >
            > PS: Martin, is there a list of contributers somewhere? There seem to
            > have been several contributers by now, especially for output filters.
            > The names I found by looking at a few files were in no particular order
            > and probably incomplete:
            >
            > Kevin O' Gorman
            > Reini Urban
            > Ian MacPhedran
            > Wolfgang Glunz
            > Masatake YAMATO
            > Bernhard Herzog
            >
            > Here the patch:
            >
            > *** output-eps.c.~1~ Tue Sep 19 20:19:04 2000
            > --- output-eps.c Fri Sep 22 16:10:06 2000
            > ***************
            > *** 122,127 ****
            > --- 122,128 ----
            > out_splines (FILE * ps_file, spline_list_array_type shape)
            > {
            > unsigned this_list;
            > + color_type last_color;
            >
            > for (this_list = 0; this_list < SPLINE_LIST_ARRAY_LENGTH (shape);
            > this_list++)
            > ***************
            > *** 132,150 ****
            > spline_list_type list = SPLINE_LIST_ARRAY_ELT (shape, this_list);
            > spline_type first = SPLINE_LIST_ELT (list, 0);
            >
            > ! c = k = 255 - list.color.r;
            > ! m = 255 - list.color.g;
            > ! if (m < k)
            > ! k = m;
            > ! y = 255 - list.color.b;
            > ! if (y < k)
            > ! k = y;
            > ! c -= k;
            > ! m -= k;
            > ! y -= k;
            > ! /* symbol k is used for CorelDraw 3/4 compatibility */
            > ! OUT4 ("%f %f %f %f k\n", (double) c/255.0,
            > ! (double) m/255.0,(double) y/255.0, (double) k/255.0);
            > OUT_COMMAND2 (START_POINT (first).x, START_POINT (first).y, "m");
            >
            > for (this_spline = 0; this_spline < SPLINE_LIST_LENGTH (list);
            > --- 133,156 ----
            > spline_list_type list = SPLINE_LIST_ARRAY_ELT (shape, this_list);
            > spline_type first = SPLINE_LIST_ELT (list, 0);
            >
            > ! if (this_list == 0 || !COLOR_EQUAL(list.color, last_color))
            > ! {
            > ! if (this_list > 0)
            > ! OUT_LINE ("f");
            > ! c = k = 255 - list.color.r;
            > ! m = 255 - list.color.g;
            > ! if (m < k)
            > ! k = m;
            > ! y = 255 - list.color.b;
            > ! if (y < k)
            > ! k = y;
            > ! c -= k;
            > ! m -= k;
            > ! y -= k;
            > ! /* symbol k is used for CorelDraw 3/4 compatibility */
            > ! OUT4 ("%f %f %f %f k\n", (double) c/255.0,
            > ! (double) m/255.0,(double) y/255.0, (double) k/255.0);
            > ! }
            > OUT_COMMAND2 (START_POINT (first).x, START_POINT (first).y, "m");
            >
            > for (this_spline = 0; this_spline < SPLINE_LIST_LENGTH (list);
            > ***************
            > *** 160,167 ****
            > END_POINT (s).x, END_POINT (s).y,
            > "c");
            > }
            > ! OUT_LINE ("f");
            > }
            > }
            >
            >
            > --- 166,175 ----
            > END_POINT (s).x, END_POINT (s).y,
            > "c");
            > }
            > ! last_color = list.color;
            > }
            > + if (SPLINE_LIST_ARRAY_LENGTH (shape) > 0)
            > + OUT_LINE ("f");
            > }
            >
            >
            >
            > --
            > Bernhard Herzog | Sketch, a drawing program for Unix
            > herzog@... | http://sketch.sourceforge.net/
          • Ted Harding
            ... Great! This now works (at least in black and white). Many thanks, also for the explanations. Ted. ... E-Mail: (Ted Harding)
            Message 5 of 20 , Sep 22, 2000
            • 0 Attachment
              On 22-Sep-00 Bernhard Herzog wrote:
              > The best solution would be for autotrace to mark the splines that
              > should be filled together. As a workaround it might be possible to
              > just treat consecutive splines with the same fill color as subpath of
              > the same postscript path and only start a new postscript path if the
              > color changes. For SVG and SK simillar solutions should work.
              >
              > I've included a patch for the EPS code below.

              Great! This now works (at least in black and white).
              Many thanks, also for the explanations.
              Ted.

              --------------------------------------------------------------------
              E-Mail: (Ted Harding) <Ted.Harding@...>
              Fax-to-email: +44 (0)870 284 7749
              Date: 22-Sep-00 Time: 17:43:50
              ------------------------------ XFMail ------------------------------
            • Bernhard Herzog
              ... The patch I posted contains a bug. The same code is used for EPS and AI and my patch doesn t generate a correct AI file, although it s a correct EPS file
              Message 6 of 20 , Sep 22, 2000
              • 0 Attachment
                (Ted Harding) <Ted.Harding@...> writes:

                >
                > On 22-Sep-00 Bernhard Herzog wrote:
                > > The best solution would be for autotrace to mark the splines that
                > > should be filled together. As a workaround it might be possible to
                > > just treat consecutive splines with the same fill color as subpath of
                > > the same postscript path and only start a new postscript path if the
                > > color changes. For SVG and SK simillar solutions should work.
                > >
                > > I've included a patch for the EPS code below.
                >
                > Great! This now works (at least in black and white).
                > Many thanks, also for the explanations.
                > Ted.

                The patch I posted contains a bug. The same code is used for EPS and AI
                and my patch doesn't generate a correct AI file, although it's a correct
                EPS file (AFAICT, anyway).

                Here's a bette one, again against the original code of Autotrace 0.24a:

                *** output-eps.c.~1~ Tue Sep 19 20:19:04 2000
                --- output-eps.c Fri Sep 22 19:57:41 2000
                ***************
                *** 122,127 ****
                --- 122,128 ----
                out_splines (FILE * ps_file, spline_list_array_type shape)
                {
                unsigned this_list;
                + color_type last_color;

                for (this_list = 0; this_list < SPLINE_LIST_ARRAY_LENGTH (shape);
                this_list++)
                ***************
                *** 132,150 ****
                spline_list_type list = SPLINE_LIST_ARRAY_ELT (shape, this_list);
                spline_type first = SPLINE_LIST_ELT (list, 0);

                ! c = k = 255 - list.color.r;
                ! m = 255 - list.color.g;
                ! if (m < k)
                ! k = m;
                ! y = 255 - list.color.b;
                ! if (y < k)
                ! k = y;
                ! c -= k;
                ! m -= k;
                ! y -= k;
                ! /* symbol k is used for CorelDraw 3/4 compatibility */
                ! OUT4 ("%f %f %f %f k\n", (double) c/255.0,
                ! (double) m/255.0,(double) y/255.0, (double) k/255.0);
                OUT_COMMAND2 (START_POINT (first).x, START_POINT (first).y, "m");

                for (this_spline = 0; this_spline < SPLINE_LIST_LENGTH (list);
                --- 133,157 ----
                spline_list_type list = SPLINE_LIST_ARRAY_ELT (shape, this_list);
                spline_type first = SPLINE_LIST_ELT (list, 0);

                ! if (this_list == 0 || !COLOR_EQUAL(list.color, last_color))
                ! {
                ! if (this_list > 0)
                ! OUT_LINE ("*U");
                ! c = k = 255 - list.color.r;
                ! m = 255 - list.color.g;
                ! if (m < k)
                ! k = m;
                ! y = 255 - list.color.b;
                ! if (y < k)
                ! k = y;
                ! c -= k;
                ! m -= k;
                ! y -= k;
                ! /* symbol k is used for CorelDraw 3/4 compatibility */
                ! OUT4 ("%f %f %f %f k\n", (double) c/255.0,
                ! (double) m/255.0,(double) y/255.0, (double) k/255.0);
                ! OUT_LINE ("*u");
                ! }
                OUT_COMMAND2 (START_POINT (first).x, START_POINT (first).y, "m");

                for (this_spline = 0; this_spline < SPLINE_LIST_LENGTH (list);
                ***************
                *** 161,167 ****
                --- 168,177 ----
                "c");
                }
                OUT_LINE ("f");
                + last_color = list.color;
                }
                + if (SPLINE_LIST_ARRAY_LENGTH (shape) > 0)
                + OUT_LINE ("*U");
                }




                --
                Bernhard Herzog | Sketch, a drawing program for Unix
                herzog@... | http://sketch.sourceforge.net/
              • Masatake YAMATO
                ... I ve applied this patch to the main source trunk. Masatake
                Message 7 of 20 , Sep 25, 2000
                • 0 Attachment
                  Bernhard Herzog wrote:
                  > The patch I posted contains a bug. The same code is used for EPS and AI
                  > and my patch doesn't generate a correct AI file, although it's a correct
                  > EPS file (AFAICT, anyway).

                  I've applied this patch to the main source trunk.

                  Masatake
                • wgibbons@bigfoot.com
                  ... and AI ... correct ... 0.24a: Thanks very much for your patch. Unfortunately I have a problem with anti-aliased bitmaps. Your patch cures the holes
                  Message 8 of 20 , Sep 26, 2000
                  • 0 Attachment
                    --- In autotrace@egroups.com, Bernhard Herzog <herzog@o...> wrote:

                    > The patch I posted contains a bug. The same code is used for EPS
                    and AI
                    > and my patch doesn't generate a correct AI file, although it's a
                    correct
                    > EPS file (AFAICT, anyway).
                    >
                    > Here's a bette one, again against the original code of Autotrace
                    0.24a:

                    Thanks very much for your patch. Unfortunately I have a problem with
                    anti-aliased bitmaps.

                    Your patch cures the 'holes' problem for non anti-aliased bitmaps I
                    have, but if I try to give AutoTrace an anti-aliased bitmap (eg of
                    the letters 'Aa') I still get the holes filled in. Although I can
                    see they grey bits aound the edges of the letters so I can actually
                    see where the holes should be.

                    Regards
                    William Gibbons
                  • Ted Harding
                    ... You might like to try a pre-filtering with ImageMagick which I have found useful. Open the file with display ; in the Files menu select Effects and then
                    Message 9 of 20 , Sep 26, 2000
                    • 0 Attachment
                      On 26-Sep-00 wgibbons@... wrote:
                      > Thanks very much for your patch. Unfortunately I have a problem with
                      > anti-aliased bitmaps.
                      >
                      > Your patch cures the 'holes' problem for non anti-aliased bitmaps I
                      > have, but if I try to give AutoTrace an anti-aliased bitmap (eg of
                      > the letters 'Aa') I still get the holes filled in. Although I can
                      > see they grey bits aound the edges of the letters so I can actually
                      > see where the holes should be.

                      You might like to try a pre-filtering with ImageMagick which I have
                      found useful.

                      Open the file with "display"; in the Files menu select Effects and
                      then Threshold. The default value of Threshold is 128; select a smaller
                      value experimentally (I usually end up with about 60).

                      I have used this for "noisy" scanned B&W images (with faint mottled
                      colours in the background), and it works well when the result is
                      submitted to autotrace. I reckon antialiasing is also noise in this
                      context!

                      Ted.

                      --------------------------------------------------------------------
                      E-Mail: (Ted Harding) <Ted.Harding@...>
                      Fax-to-email: +44 (0)870 284 7749
                      Date: 26-Sep-00 Time: 19:15:26
                      ------------------------------ XFMail ------------------------------
                    • Bernhard Herzog
                      ... Well, the patch is a bit of a kludge. The real problem, AFAICT, is that the list of splines that autotrace creates and passes to the export filters doen t
                      Message 10 of 20 , Sep 26, 2000
                      • 0 Attachment
                        wgibbons@... writes:

                        > Your patch cures the 'holes' problem for non anti-aliased bitmaps I
                        > have, but if I try to give AutoTrace an anti-aliased bitmap (eg of
                        > the letters 'Aa') I still get the holes filled in. Although I can
                        > see they grey bits aound the edges of the letters so I can actually
                        > see where the holes should be.

                        Well, the patch is a bit of a kludge. The real problem, AFAICT, is that
                        the list of splines that autotrace creates and passes to the export
                        filters doen't contain enough information to make a correct postscript
                        output in all cases. Could you send me (by private mail) a sample image
                        that shows the problem? I'd like to find out what's going on.

                        --
                        Bernhard Herzog | Sketch, a drawing program for Unix
                        herzog@... | http://sketch.sourceforge.net/
                      • William Gibbons
                        Hi, You asked for a file which exhibits this problem, so I m sending you a .BMP file. I don t think there s anything particularly special about this file,
                        Message 11 of 20 , Sep 26, 2000
                        • 0 Attachment
                          Hi,

                          You asked for a file which exhibits this problem, so I'm sending you a .BMP
                          file. I don't think there's anything particularly special about this file,
                          I've had the same problem with other similar files.

                          Thanks for your speedy response.

                          Regards
                          William

                          ----- Original Message -----
                          From: Bernhard Herzog <herzog@...>
                          To: <autotrace@egroups.com>
                          Sent: Tuesday, September 26, 2000 10:00 PM
                          Subject: Re: [AutoTrace] Re: autotrace 0.23


                          > wgibbons@... writes:
                          >
                          > > Your patch cures the 'holes' problem for non anti-aliased bitmaps I
                          > > have, but if I try to give AutoTrace an anti-aliased bitmap (eg of
                          > > the letters 'Aa') I still get the holes filled in. Although I can
                          > > see they grey bits aound the edges of the letters so I can actually
                          > > see where the holes should be.
                          >
                          > Well, the patch is a bit of a kludge. The real problem, AFAICT, is that
                          > the list of splines that autotrace creates and passes to the export
                          > filters doen't contain enough information to make a correct postscript
                          > output in all cases. Could you send me (by private mail) a sample image
                          > that shows the problem? I'd like to find out what's going on.
                          >
                          > --
                          > Bernhard Herzog | Sketch, a drawing program for Unix
                          > herzog@... | http://sketch.sourceforge.net/
                          >
                          >
                          >
                          >
                        • wgibbons@bigfoot.com
                          ... Thanks for the suggestion. I don t believe the anti-aliasing is directly responsible for the problem - it s a pretty large hole - but indirectly, perhaps.
                          Message 12 of 20 , Sep 26, 2000
                          • 0 Attachment
                            --- In autotrace@egroups.com, (Ted Harding) <Ted.Harding@N...> wrote:
                            > I have used this for "noisy" scanned B&W images (with faint mottled
                            > colours in the background), and it works well when the result is
                            > submitted to autotrace. I reckon antialiasing is also noise in this
                            > context!

                            Thanks for the suggestion. I don't believe the anti-aliasing is
                            directly responsible for the problem - it's a pretty large hole - but
                            indirectly, perhaps.

                            As for anti-aliasing being noise, well...anti-aliasing is a way of
                            simulating a higher resolution display than you actually have. If
                            it's done properly, you can send more information about the paths
                            than you can with the same resolution monochrome bitmap, because (for
                            instance) a mid-grey pixel indicates that a pixel should be half-
                            covered by a path, and you can't indicate this with monochrome (or
                            non-anti-aliased bitmaps).

                            However, it may well be that trying to extract this information and
                            use it to improve curve tracing would be more trouble than it's
                            worth. I don't think it's a trivial problem. So yes, perhaps the
                            best solution is to remove anti-aliasing with a noise filter.
                            However, I'm not sure this helps if you're trying to trace
                            photographs.
                          • wgibbons@bigfoot.com
                            ... [ 23K binary included ] Sorry about this, I didn t mean to send this mail to the discussion... William
                            Message 13 of 20 , Sep 26, 2000
                            • 0 Attachment
                              --- In autotrace@egroups.com, "William Gibbons" <wgibbons@b...> wrote:

                              [ 23K binary included ]

                              Sorry about this, I didn't mean to send this mail to the discussion...

                              William
                            • Ted Harding
                              ... Attached are two EPS files: text.eps is the autotrace conversion (using 0.24a with patched output-eps.c) of William Gibbons s original file text.bmp
                              Message 14 of 20 , Sep 27, 2000
                              • 0 Attachment
                                On 26-Sep-00 William Gibbons wrote:
                                > Hi,
                                >
                                > You asked for a file which exhibits this problem, so I'm sending you a
                                > .BMP file. I don't think there's anything particularly special about
                                > this file, I've had the same problem with other similar files.

                                Attached are two EPS files:

                                text.eps is the autotrace conversion (using 0.24a with patched
                                output-eps.c) of William Gibbons's original file text.bmp

                                text2.eps is the result of running autotrace on a BMP file obtained
                                by running the noise-filtering I described previously (using the
                                "display" program from ImageMagick and lowering the threshold
                                -- in this case to a hit-or-miss value of 64).

                                The hit-or-miss seems to have hit it spot on.

                                Note also the difference in file size ... !

                                Ted.

                                --------------------------------------------------------------------
                                E-Mail: (Ted Harding) <Ted.Harding@...>
                                Fax-to-email: +44 (0)870 284 7749
                                Date: 27-Sep-00 Time: 15:08:28
                                ------------------------------ XFMail ------------------------------
                              • Stefan A. Deutscher
                                ... Hi Ted, hi all, can we keep (binary) attachments down to a minimum, please? Sometimes direct exchange between the involved folks might be enough. The
                                Message 15 of 20 , Sep 27, 2000
                                • 0 Attachment
                                  On Wed, Sep 27, 2000 at 03:08:29PM +0100, Ted Harding wrote:
                                  > On 26-Sep-00 William Gibbons wrote:
                                  > > You asked for a file which exhibits this problem, so I'm sending you
                                  > > a .BMP file. I don't think there's anything particularly special
                                  > > about this file, I've had the same problem with other similar files.
                                  >
                                  > Attached are two EPS files: text.eps is the autotrace conversion
                                  > (using 0.24a with patched output-eps.c) of William Gibbons's original
                                  > file text.bmp text2.eps is the result of running autotrace on a BMP
                                  > file obtained by running the noise-filtering I described previously
                                  > (using the "display" program from ImageMagick and lowering the
                                  > threshold -- in this case to a hit-or-miss value of 64). The
                                  > hit-or-miss seems to have hit it spot on. Note also the difference in
                                  > file size ... ! Ted.

                                  Hi Ted, hi all,

                                  can we keep (binary) attachments down to a minimum, please? Sometimes
                                  direct exchange between the involved folks might be enough. The problem
                                  is that I, for one, am accessing my email over a 14.4k dial-up modem and
                                  in this part of the world even local calls are metered to be paid.

                                  Thanks & Cheers,

                                  Stefan


                                  --
                                  ============================================================================
                                  Stefan A. Deutscher |
                                  Donostia International Physics Center | office: ++34-943-018174
                                  Universidad del Pais Vasco, Facultad de Quimica | fax : ++34-943-015600
                                  Departamento de Fisica de Materiales | home : ++34-943-270647
                                  Apartado 1072, San Sebastian 20080, Spain | email : sad@...
                                  ============================================================================
                                  The early bird gets the worm but the second mouse gets the cheese.
                                • William Gibbons
                                  ... Yes. One would expect this given the way that AutoTrace seems to handle the tracing of anti-aliased bitmaps. If I had simply wanted to get a more
                                  Message 16 of 20 , Sep 27, 2000
                                  • 0 Attachment
                                    Ted Harding wrote:

                                    > Note also the difference in file size ...

                                    Yes. One would expect this given the way that AutoTrace seems to handle the
                                    tracing of anti-aliased bitmaps. If I had simply wanted to get a more
                                    accurate trace of the characters, I would have chosen to export with
                                    anti-aliasing turned off. I was really just trying to report what I believe
                                    to be a bug - the fact that the holes are filled in shouldn't happen,
                                    regardless of the anti-aliasing.

                                    Some other bitmap trace algorithms perform differently with anti-aliased
                                    bitmaps, even with noise filters turned off. As I said, I believe there are
                                    better ways to deal with anti-aliased bitmaps than simply treating
                                    anti-aliasing as noise, and indeed dithering is a fact of life for tracing
                                    photographs etc.

                                    William
                                  Your message has been successfully submitted and would be delivered to recipients shortly.