Member Provanum Group

Raster to vector 'how to'

Here are links to our conversion services and software.

Raster to vector conversion is the process of converting images from scanners, photographs or bitmap generating applications to editable vector formats. CAD users need editable vectors because CAD applications (such as AutoCAD, Solidworks, TurboCAD, more) use vectors. And graphic designers may need vectors if they are designing in an application such as Illustrator or Freehand.

In this article we are describing the process from the point of view of a CAD user who needs the images for engineering design. But the principles can also be applied for graphic design.

What's the difference between a raster and a vector? Take a look at the illustration below.

If you open a vector image and zoom in on a detail the lines will appear fine no matter how far in you zoom. If you open a raster image and zoom in on a detail the lines will appear jaggy.

The raster image appears jaggy because the lines have been created, in the case of a scanner, by taking individual snapshots of 'dots' on the original sheet. When the dots are put together on a screen our brains literally 'connect the dots' and we see lines.

This is fine for viewing scanned images. Provided the original raster image is of a sufficiently high resolution - i.e. there are plenty of dots per inch (dpi) - our brains will do the connecting work. 

CAD programs work completely differently. They describe lines mathematically. A CAD program defines a straight line as "start at coordinate x₁,y₁ and go to coordinate x₂,y₂.

Curved lines are described using more complex mathematical descriptions.

The process of converting from the raster dots to the mathematical vectors is called raster to vector conversion.

There are essentially three approaches:

• Automatic Conversion Software - traces the dots in the raster image and automatically replaces the dots with CAD vector lines.
  
• Manual Underlay Conversion - insert the raster image in a CAD application and draw vectors over it in CAD.
  
• Manual Digitizing - use the mouse or a digitizing tablet to manually trace over each original line.

Why would one even consider a manual approach when an automatic option is available? Well, even though the initial vectorization is automatic, there is almost invariably some manual reworking required to clean up the vectors created by automatic vectorization software.

And, as a rule of thumb, the worse the quality of the original raster image (degradation by fading, coffee stains, etc.) and the greater the clutter and overprinting in the drawing, the more manual work will be required. We've tried to illustrate this in the diagram to the right.

So why use automatic software if some of the work will have to be done manually? Say you have a drawing which would take 10 hours to redraw using a completely manual approach and an automatic vectorization software that converts the same image (in a matter of seconds) but then requires 5 hours of manual clean-up. Which would you choose?

Why can't automatic conversion software do a perfect job?

All the computing power of the modern personal computer can't come near the human vision. So it's easy for you to look at a detail on an image and know that it's a 'joist hanger detail' or 'tapered sleeve' or whatever. You can do this because your brain interprets what it sees in the context of the overall drawing.

However the limited power of the computer just sees 'a lot of lines coming together'. It probably won't draw the junctions anything like as pristinely as you would draft them in CAD. And it will usually be very literal - if a line is in the scan the vectorizing software will replicate it, whether it should be there or not.

When will automatic vectorizing software do a perfect job? And when should it be avoided altogether?

Logos and artwork scanned for NC cutting can often be converted and used with either no or only minimal manual clean-up. Clean architectural or mechanical sheet drawings work well.

At the other extreme, the coffee-stained sheet recovered from the shop floor, barely held together with tape, will not scan well. Expect this to be a hard scan for a human to interpret, let alone automatic conversion software. Ancient vellums and faded blues will create similar problems.

So should I buy software or have someone do conversions for me? 

How is text converted?

If you are converting to vectors for CAD use text should be converted to real characters. This means an 'A' is stored as the character 'A'. However if you are converting to vectors for use in cutting software you'll want an 'A' to be converted to vector lines - two sides and a horizontal.

Automatic conversion software is usually equipped with an OCR function to convert raster text to characters. If this is not used the text is converted to vector lines suitable for use in cutting applications.

If you outsource the conversion be sure to tell the converter whether you need character or vector line text. If you specify vector line text you should also specify if you want center-line or outline vectors. 

I have large prints but no large format scanner.

Large format scanners are expensive. Expect to pay thousands of dollars if you are planning a purchase (and take a look at our notes on the subject).

Increasingly large format scanners are found in printing/copying centers. But beware. Scanning competences vary widely. If your prints are pristine, black ink only, you should ask the operator to scan to a 1-bit image with no dithering.

Dithering changes the original image so it looks prettier to the eye but it actually worsens the conversion results (more on dithering).

Then stay with the operator to make sure, while you are still in the store:

i) That the file format is 1-bit black and white (we suggest you specify TIFF CCITT Group 4 file format as it's only 1-bit and is almost universally readable).

ii) That the final image is actually legible to you when you zoom in on details. If it's not legible to you it won't be legible for conversion. This can often be corrected by increasing scan resolution for more dots per inch.

Quite often operators will just use the default settings on the scanner and you end up with a multi megabyte color file when you could have had just a couple of hundred KB black and white file.

If you are outsourcing conversion work and your original prints are at all degraded - visually or physically - ask the conversion service to include scanning in their quote. Their scanning experience will enable them to create useable scans and not damage your originals in the process.

A reputable converter will be able to quote you if you first send them digital camera shots of the original prints - they should not need the originals in order to prepare a quote.

Avoid scanning to JPEG format if you possibly can.

We suggested TIFF CCITT Group 4 file format for scanning black and white originals. If your original is in color you'll have to use a color format. You should avoid JPEG since this is a 'lossy' format. JPEG drops some colors in order to create a smaller file. This is very clever and useful for photos but results in an image which will not convert as well to vectors.

Beware the fixed price conversion offer.

Unless you're converting a very common image, such as that of a national flag, you can be pretty sure your conversion needs will be unique. When you request quotes it's reasonable for you to be given an indication of price based on your description. But a reputable conversion service will want to see scans or photos of the originals before finalizing a quote.

If you are offered a fixed price you can be pretty sure your job will not be given individualized attention.

Where can I find conversion services or software?

We at Trix Systems offer:

Conversion Services

Software

Your needs are unique and we're pleased to discuss them with you.

Call us on (978) 256-4445.