We know that the human eye can identify thousands of patterns at a glance. For example, 1000 patterns can be identified intuitively. If we take two steps (using hierarchy), we can recognize 1000000=1,000,000 patterns. The hierarchy concept, thought unique to human beings, is an excellent way to recognize a multitude of objects. This concept also supports higher level communication with only small amounts of information.
Fig.1 Input image signal (uppermost) into the eye is processed through the retina and cortex toward higher levels.
Let's look at human vision more closely. An object recognized by the eye comprises many components, called sub-objects or primitives. Because each primitive is composed of pixels, human vision is organized into a hierarchical architecture. Figure 1 shows the hierarchical change that information undergoes from retina to cortex, where the volume of the processing signal is reduced along the way. (Cheng, Ewe, Chiu, Bashir, J. Micromechanics and Microengineering, 11(2001), 487-498)
Compare the two photos in Figure 2. If you have small knowledge about art, you will easily see the significant difference between the two. Eye size differs only slightly, but the perceived difference is great even though more than 99% of the pixels are identical. If you are unfamiliar with this face, the perceived difference will be smaller.
Fig.2 Human perception is sensitive to the human face. Even a slight change has a great effect, but it depend on the object to which it belongs.
(Leonardo Da Vinci, Harry H. Abrams, Inc. 2000)
This simple experiment suggests that visual objects are hierarchically organized with differences in the contribution ratio. It is well known that even macaques (like the Japanese monkey) have the ability to recognize the faces of their own species, which indicates that macaques have the same layered hierarchical perception.
In the modern office, most business documents are hierarchically organized. Letters and figures are higher layered images composed of lower layer primitives: color, edges and lines. When "reading" the document, we focus on upper layer words and sentences while ignoring background color or noise.
Now look at the real world. The world, however, is made up of unique individuals. Those who are sensitive to backgrounds should have different opinions from the majority, who may be sensitive to foregrounds. Our most recent technolog to reduce "show-through effect," where images show-through from the reverse side of the paper, may satisfy these unique individuals. (H. Nishida and T. Suzuki, "A multi-scale approach to restoring scanned color document images with show-through effects," in Proceedings, Seventh International Conference on Document Analysis and Recognition (Edinburgh, UK), August 3-6, 2003, pp. 584-588)