You are here: Technology Digitization Digitization of dotted lines

Digitization of dotted lines

Rate this item
(0 votes)

Until recently, the dotted lines in Easy Trace Pro can only be vectorized in manual or semiautomatic mode. However, they are quite amenable to automated digitization.

Only manual and semiautomatic modes were available for vectorizing of dotted lines in Easy Trace till recently. Still, they are quite suitable for automatic tracing. The video Vectorizing of dotted lines in Easy Trace Pro" demonstrates the process.

Tracing method consists of the following steps.

  1. extraction of the specialized "dot image";
  2. dot contouring;
  3. forming of circles out of dot contours;
  4. transformation of the circles into short hachures;
  5. joining of the hachures as line segments nonmetering their directions.

The very name of the first step indicates its main task - to preserve dots at extraction of the image. line quality is of no importance. There are some niceties though

  • First, the higher is resolution of the image (DPI value), the more dots approximate to circles.
  • Second, many dots are "glued" to grid lines. It is necessary hence to vectorize the grid first and delete it from the image with the help of the Rasterize Vector Data utility.
  • Third, the following operations should be performed at the final stage of black-and-white dot image extraction:
    1. "Diffuse" - radius is one half of the typical dot diameter, factor is 1.5. The operation smoothes dot boundaries, makes dot shape more round;
    2. "Unsharp Mask" - radius is one half of the typical dot diameter, factor up to 1000. The operation "unsticks" dots from lines and each other;
    3. "Subject Layer Extraction" - specify Admissible Brightness Range for final separation of the dots from the background.
    At the second step (automatic outlining), save "Only without holes" contours. We do not want to vectorize "O" letters , zeros, etc.

    At circle autodetection (the third stage), remember that superfluous dots are much more "harmful" than omitted ones. Be careful at selection of the range of diameters and deviation from the ideal round shape.

  • Fourth step - Object Conversion. Transform circles into strokes 2-3 pix long to enable applying of the "Treat lines as points if their length is shorter than..." option of the Breakups Joining utility.

  • Fifth stage - specifying of joining parameters that depend on the character of dot lines. In rare instances when there is a lot of vector "rubbish" together with vectorized dots, breakup joining may be subdivided into 2 steps:
    1. Join most reliable parts of lines first. Specify the radius to search line continuation within just a little more than distance between neighboring dots; minimal angle of line convergence should be about 140 degrees;
    2. Use the Eraser Tool to delete obvious mistakes (false dots, incorrect junctions in "bottlenecks");
    3. Perform final breakup joining. The distance for search of continuation is about 2.5 dot-to-dot intervals at a negative convergence angle ( -20 degrees).

For previous versions

The following algorithm of dots-to-strokes conversion may be recommended:

  1. Extract the dot image with compulsory diffusion and brightness filtering to delete numerous irrelevant small objects;
  2. Image thinning. The dots turn into small islands made of 1-5 pix;
  3. Apply Mask Filtering (the "Strokes from dots" strategy) to transform the dots into strokes;
  4. Automatic Tracing -> Autotrace lines;
  5. Apply the "Select objects by a query" command of the Group Editor to select and delete all line segments longer than 3 vertices.

Subsequent joining of the strokes into lines agrees with operations described above for up-to-date versions.

The "old" method of dotted line autotracing has some disadvantages yet. Rather large objects may turn into dots at thinning if they are round. In turn, mask filtering makes strokes of ANY pixel conglomerate if it finds room within the 3X3 mask. It results in superfluous points and thus prolongs further editing.

The recommendation on two-stage breakup joining with error correction is rooted in our "Murmansk Region " project. We had to process overcompressed JPEG images where all crossings of rivers and isolines (red + blue = black) generated false dots.

The method we suggest now is founded on dot extraction as round objects having diameters within the specified range. It means selection subject to both shape (round/ non-round) and diameter (too large/ too small) that ensures high reliability of recognition.

We manage to extract dotted lines to within 90-95% even from dirt gray-scale images of rather low resolution.

More in this category: « Prev Next »

Add comment


Security code
Refresh