• Single Shapes
  Each color layer represents individual color shapes
• Combine Shapes
  Each color represents one layer. Shapes on upper layers drop a "shadows" on all layers below, in the color of the actual lower layer. This can reduce gaps between shapes, because the color of the lower layers shine through
• blur
  Duplicates all shapes with some blur radius and places them in the background
• fill
  Same as "merge", but the "shadow" on the lower levels is feathered at the edges to prevent shine through at shape edges. If feathering removes too much small detail shapes from your input image, try "merge" background.

The blur parameter defines the smoothing factor around each color patch.

• How to remove white (or transparent) empty spaces (or lines) around color patches?
  • Increase blur factor until all empty spaces are gone

Select a filter to reduce the number of colors in an image

• None: disable filtering
• Canny (experimental): Preprocess the input image with Canny Filter
• Remove Background: This filter tries to find the dominant background color from the colors on the borders and the removes pixels with silimar colors using flood fill algorithm.
• Remove Color: Remove all pixels with similar color to the specified color.

Configure the layers (areas of same colors) for the output vector image:

• Ignore one color for the output?
  • Move the mouse over the color circle and a disable the checkbox
• Too many small speckles around larger contours?
  • Move the color circle of the larger contour to the left and the color circles of the speckles to the right
• Group similar colors automatically?
  • Use the slider or the 'tolerance' parameter
• Remove a color from a group?
  • Drag & drop the color circle in the 'free' (gray) area of the layers panel and release the mouse button

The mins parameter defines the size of speckles which will be ignored during vectorization.

• Too many small speckles around larger contours?
  • Increase the mins parameter

For small input images with fine features, like text or small shapes, sometimes the vectorized output does not show all required image features.
With the help of artificial intelligence (AI), the input image can be upscaled before vectorization to help the vectorization algorithms to detect the vector shapes better.
Upscaling AI is optimized for low color count clipart style images.

Input images including pixel noise, for example from JPEG compression artifacts sometimes produce to complicated vector outputs, because the noisy pixels are also detected as relevant vector shapes.
With the help of artificial intelligence (AI), input pixel noise can be reduced in different levels from low to highest.

Select the smallest amount of colors suitable for your image for best performance and smaller output vector file (SVG).

• Some colors not shown correcty?
  • Increase the colors one step at a time until all colors look OK in the left preview image

Number of colors in vector output.

enable crowd source and your image will be also edited by other users of www.vectorizer.io and you will see their edits in the "variants" section.

Threshold defines the factor which is used to automatically merge similar colors in the layers list.

• How to reduce the numbers of colors in the layers list?
  • Increase the tolerance until similiar colors are merged together in groups
• How to increase the detail level in the output image?
  • Decrease the threshold until the output image has the detail level you want