carol.gimp.org

gimp basics

another background removal tutorial

I was making an index page for my writing section and I was using a handful of photos that already had dark backgrounds. These images are sort of special because with a little work they look good yet without the work they (usually) don't look that good.

I made a gallery of images from my first year with my camera of images that might be nicer with the dark backgrounds cleaned up and made darker, feel free to use an image from the dark images set to follow the tutorial with.

The original image can be found in that collection img_0317 to try some of this with.

candles

The Decompose Filter

Decompose is found in Gimp2 by right clicking on the image and following the menus through Filters -->Colors -->Decompose.

This plug-in pops up a dialog that is ready to decompose the active image layer into a layered grayscale image that represents the amount of Red, Green and Blue in the original layer.

While there are many different options to decompose your image with; for the purposes of this tutorial and the images I was working with for the gallery, I used the default decompostion dialog setting (RGB) and also (as pictured) the HSV option (Hue, Saturation and Value).

screen-decompose-HSV

Decomposition Layers

The amount of white on each layer is the amount of that color property in the image. The layers are labeled in the layer comments with the name of the image color property the decomposition represents.

Red, Green and Blue Decomposition

red

red

green

green

blue

blue

Hue, Value and Saturation Decompostion

hue

hue

saturation

saturation

value

value

Making a Mask

The goal right now is to make a mask quickly and efficiently, so when choosing which decomposition to use, keep in mind that the white will show the image and the black will mask the original. When making the ten images for the dark images gallery, depending on the image, I used one of the red, green or blue layers or the value layer. For this image the value layer seemed to be the best representation of what I wanted to keep.

screen-threshold

Threshold Tool

The threshold tool can be found by choosing the Threshold Tool from the toolbox or by right clicking on the image and following the menus through Layers -->Colors -->Threshold.

Simply clicking this tool on the image face will reduce the colors to two: black and white. Change the levels of threshold by either changing the numbers in the boxes located under the histogram graph of the image or by clicking with the mouse in the histogram portion at the beginning and end of the area. (Truthfully, I don't understand the tool well enough to write about it here and I almost always use the default setting.)

candles-value-threshold

Work on the mask

The next steps are applied directly to the layer mask. You should be able to see the changes to the mask by what is showing of the image underneath the mask. Putting a black background underneath the masked image will be helpful also.

Add the decomposed image as a mask

Add the decomposed and thresholded image to the original image as a layer mask.

Use a Blur Filter

The thresholded value image is jaggy so the first task is to soften the edges by using one of the Gaussian Blur Filters found by right clicking on the image and following the menus menus through Filters -->Blur -->Gaussian Blur

I blurred for a 30 pixel radius for this image.

Use the Levels Tool

The Levels Tool found on the toolbox or by following the menus through >Image< -->Layers -->Colors -->Levels. Use the center triangle on the histogram to make the white area bigger by moving it to the left.

mask1

Add the mask.

masked1

The resulting image.

mask2

Blurred mask.

masked2

The resulting image.

mask3

Expanded white area.

masked3

The resulting image.

Clean Up

Once I was happy with the parts of the image showing through the mask that I liked, I simply had to paint with black on the mask over the parts I did not want to show.

Other Background Manipulation Tutorials

another-gnu-type-sm

Most all of my software is gnu. thanks!

Creative Commons License

This work is licensed under a Creative Commons License.