Tutorial: Vectorize Scans
This tutorial describes how to convert a scanned image into vector art. This process is called tracing or vectorization, and can either be done manually or using an automatic tool. In this tutorial, we explain how to use Vector Magic to perform this conversion quickly and easily.
What kind of images does this tutorial apply to?
This tutorial does apply to:
- Scans of artwork that was originally drawn or printed on paper
It does not apply to:
- Bitmaps that were digitally rasterized from a vector art original. See separate tutorial
- Photographs. See separate tutorial
- Images drawn in a bitmap editing tool like Photoshop.
What do I need to follow this tutorial?
-
This tutorial uses Vector Magic Desktop Edition. You can download a trial version from the desktop application page.
-
The Online Edition is very similar to the Fully Automatic mode in the Desktop Edition. The user interface looks slightly different, and there's no transparency support. With those caveats in mind, you can also use the Online Edition to follow this tutorial.
Best practices for scanning art that will be vectorized
Scanning hardware and software varies between manufacturers, so I will keep my remarks as general as possible. The following remarks are guidelines only; please experiment with your particular hardware, software, and source material to find the best choices for your particular case. I have found:
-
For black and white or grayscale source material, choose the grayscale scanning option.
- Do not choose the black and white scan mode, as this results in loss of data through quantization.
- Do not use the color scan mode, as this leads to undesired color artifacts in the scan around shape boundaries.
- For color source materials, use the color scan setting.
- Turn off any unsharp mask or other contrast or edge-enhancing features. While such features do enhance certain aspects of the scanned result from an aesthetic perspective, they only make it harder to vectorize.
- Make sure your scan bed is clean and dust free.
- Scan at 150-300 dpi. For typical consumer-grade scanners, it rarely makes sense to scan at higher than 300 dpi, as scanners are mostly hallucinating any details at a higher resolution anyway. I would typically choose 150 dpi for most scanning, especially if you intend to vectorize the resulting bitmap.
- Do not scale, filter, or otherwise alter the raw scan before vectorizing. You may crop it to the correct size, but more advanced operations typically only serve to remove information from the scan.
-
I found an interesting public-domain vector image of a Celtic knot design. You can download the SVG version of this here. Here is a rasterized bitmap version of that file:
- Next, I printed that file using a high-quality laser printer. This resulted in a very crisp, clean print - out of the image, filling a 6 inch square.
- Then, I scanned this using a consumer-grade Canon scanner at three resolutions: 150, 300, and 600 dpi.
- Load the image by dragging and dropping it onto the application.
- Select the "Basic Wizard."
- Select "Logo with Blended Edges."
- Select "Medium" quality.
- Select "Two Colors."
- Select "Done Reviewing."
- Save the output as desired.
A closer look at scan quality at different resolutions
In order to illustrate some of the characteristics of scans, I've done a little experiment.
In this section, I will describe my findings. As I indicated in the best practices section above, I concluded that it doesn't really make any sense to scan at higher than 300 dpi. Sure, scanning at 600 dpi, gives you an image of twice the size, but it is also twice as blurry.
The scans themselves are too large to show in-line in this tutorial, but here are the links to the various sizes: 150 DPI, 300 DPI, and 600 DPI.
Zooming in on an edge in the image is instructive.
|
|
|
In the 150 dpi case, the blurring at the edge is essentially just one pixel wide. This is how it should be. If the edge is infinitely sharp, then it should only be seen by one pixel in the scanner's sensor. That pixel should then be some appropriate gray, depending on how far across the pixel the edge actually lay.
In the 300 dpi case, the blurring spans two or three pixels, indicating that some blurring has occurred. It is not clear whether this is due to limitations of the physical sensors or some processing done by the scanner or the scanner software. In either case, the effective resolution of this image is not 300 dpi.
Likewise in the 600 dpi case, where the blurring spans three or four pixels. Again, such a blurry result should be considered to have a much lower effective resolution.
In this case, I think it would make the most sense to scan at 150 dpi. In cases where a better scanner were available, it might make sense to scan at 300 dpi.
Vectorizing the scan
I have elected to use the 150 dpi scan. Using Vector Magic Desktop Edition, I vectorize it with the following steps:
Evaluating the result
By and large, the result is very good. There are a few corners that are slightly rounded in the result but that are sharp in the original, but for the most part the vectorized result matches the original quite well. The SVG output from this vectorization may be downloaded here.
The screen shots below show a zoomed in section of the result.The first image is the vectorized result, the second image is the corresponding section of the bitmap, and the third image is an overlay of the vector paths on top of the bitmap original. In this section, the result appears crisp and clean, and clearly reproduces the original shape very well.
A low quality scan
Unfortunately, not all source images are as crisp and clean as the previous example. That example was generated by printing a vector graphic with a laser printer and then immediately scanning it back in. The following example is much more challenging.
This scan is much more typical of scans. The person who scanned it made the mistake of leaving the scanner in color mode, so there are color edge artifacts along the shape boundaries. The original print quality appears to have been poor, and the shape is generally irregular, as though it were hand drawn. A size-reduced version of the image is shown below, and the full resolution version may be found here.
The following zoom view illustrates some of the challenges of this source image.
The discoloration is clearly visible. The tops of the white shapes are tinted blue and the bottoms red. The edges also appear highly irregular.
The vectorized result
Again, I use Vector Magic Desktop Edition to vectorize this image. In this case, I try both "Medium" and "Low" quality settings to see how they turn out.
It should be noted that when a large image, such as this one, is loaded into Vector Magic Desktop Edition, the program analyzes the image to determine whether it makes sense to reduce the size of the image. Please note that the resize function that is used is very carefully selected to improve the quality as much as possible through effective use of pixel averaging. Most resizing features in commercial bitmap editors do not perform scaling operations in this way, so please rely on Vector Magic to perform any shrinking operations. And never increase the size of a bitmap before loading it into Vector Magic. Such efforts only ever reduce the quality of the result.
The results are available here: low (SVG), medium (SVG).
The next three images show highly zoomed in views of the original scan and the two vectorized results. The first is the original, the second is from the low result, and the third is from the medium result.
Which of the two results is better is a matter of preference. The low result uses fewer nodes and results in a simpler geometry, while the medium reconstructs the geometry more faithfully but uses more nodes. I personally prefer the medium input quality result but I can see the merits of both.
I have included one final image of the paths from the medium result over the top of the bitmap. As usual, you can see that the paths very closely follow the source bitmap. This characteristic is one of the great strengths of Vector Magic relative to other auto-vectorization tools.
Conclusions
This tutorial has provided guidelines and advice for how to best vectorize scans using Vector Magic Desktop Edition. A number of practical tips were provided to help you get the most out of your source material and your scanner. And we showed two example of scans being vectorized: the first was a high quality scan of a crisp good-quality original, and the second was a low-quality scan of a low-quality original.
I strongly encourage you to play around with this. Vector Magic can be a powerful tool for vectorizing scans if used properly.
