Visualizing image data using shaded graphics

An alternative method for visualizing image data is presented-the representation of the luminance surface as a shaded graphic. It is shown that this representation gives insights into the structure of images, and the effects of processing, particularly for understanding feature detection. Such shaded graphics consist almost entirely of Type 1 features. Their use as test images for feature detection confirms that valley detection is appropriate for Type 1 features, when lighting is from the viewing direction. Valley detection is superior to edge detection for orthogonal lighting also, but not optimal. >