Image Processing Toolbox

Image Types

The functions that use interpolation take an argument that specifies the interpolation method. For most of these functions, the default method is nearest neighbor interpolation. This method produces acceptable results for all image types, and is the only method that is appropriate for indexed images. For intensity and RGB images, however, you should generally specify bilinear or bicubic interpolation, because these methods produce better results than nearest neighbor interpolation.

For RGB images, interpolation is performed on the red, green, and blue image planes individually.

For binary images, interpolation has effects that you should be aware of. If you use bilinear or bicubic interpolation, the computed values for the pixels in the output image will not all be 0 or 1. The effect on the resulting output image depends on the class of the input image:

• If the class of the input image is double, the output image is a grayscale image of class double. The output image is not binary, because it includes values other than 0 and 1.
• If the class of the input image is uint8, the output image is a binary image of class uint8. The interpolated pixel values are rounded off to 0 and 1 so the output image can be of class uint8.

Note    For bicubic interpolation, you may need to damp doubles to within the [0 1] range.

If you use nearest neighbor interpolation, the result is always binary, because the values of the interpolated pixels are taken directly from pixels in the input image.

Interpolation

Image Resizing