The Wikipedia example is something you probably would not really want to do, but does demonstrate just how much information you can add to a non-lossy image format without it being obvious to the human eye.
Importantly, the technique used in Wikipedia only works for hiding a lower-quality image of the same dimensions as the container in the low bits of the main image. More general steganography programs will attempt to store more arbitrary messages, and will use different storage strategies. It is unlikely that any would support the example approach for "hidden-image-in-image" shown in Wikipedia. Instead you can verify the Wikipedia example using a short program in e.g. Matlab, to mask out and re-normalise the low bits.
In pseudocode:
tree_image_pixels <- load_image( 'example.png' )
cat_pixels <- ( tree_image_pixels & 3 ) * 85
save_image_from_pixels( 'cat.png', cat_pixels )
The value of 85 is so that maximum pixel value is 3 * 85 = 255, using the full colour range. Lower values also work, but produce a less bright image.