Back

Color Channels of an RGB Image

Normally, images are thought of as being two dimensional. In truth the majority of images are actually stored as a grid of pixels where each pixel holds channels with the amount of the colors red, blue, and green to display. This means that an RGB image can be thought of as 3 dimensional with the different color channels making up the layers.

Explore the image below to see how the original image on top is broken down into its 3 color channels. Note how the white stripe on the track is clearly visible in all 3. This is because white is the result of all the colors while black is the absence of them.

This is important in convolution because it means that instead of having just a single kernel for an image there will actually be a kernel for each of those color layers. The output is also 3 dimensional unless the dimensionality is reduced through addition of the layers.