previous page| next page

Theory

Shatford extends Panofsky’s theory(see Erwin Panofsky, Studies in Iconology: Humanistic Themes in the Art of the Renaissance. New York: Harper & Row, 1962.) of three levels of meaning in a work of art to general subject analysis of pictorial work (Shatford 1986: 43). Only the first and second levels of Panofsky’s theory are to be indexed, since the third seems exemplified by the content of a subjective critical review. Jacobs interprets Panofsky’s third level as requiring extensive interpretation and cultural knowledge that is reserved for art history and similar domains. (Jacobs 120). Panofsky’s first two levels are the basis for Shatford’s generic and specific levels of description.

Table 1. Panofsky’s Levels of Meaning (Shatford 1986)[‡].

Level of Meaning	Examples	Panofsky’s term		Shatford’s term
first	Factual: man, woman, tree Expressional: grief, peacefulness, happiness, anger	pre-iconography (description) – “primary or natural subject matter” requiring “everyday familiarity with objects and events.”		generic description of the objects and actions represented. The factual are descriptions Of , expressional represent the About.
		factual	expressional
second	fat jolly man sitting in lotus position is of the Buddha about compassion.	iconography (analysis) – “secondary or conventional matter”, requiring knowledge of culture.		specific objective meaning descriptions Of and mythical abstract or symbolic About
third	Zanussi’s Illumination (film) is an instance of cinema-verite, and about science, ethics, and enlightenment. An image of a grey brick wall is about socialism.	iconology (interpretation) – “intrinsic meaning of content”, requiring synthesis, knowledge of artistic, social and cultural setting		not intended to be indexed except in highly specialized domains since agreement on meaning at this level is too difficult to achieve.

 

Shatford uses Frege’s distinction between sense and reference to show that “images may be defined as referents for the sense of the words used to describe them” (Shatford 1986: 46) A single image can be the referent for many different senses of words, or conversely and more conventionally: a picture is worth a thousand words.

The kinds of subjects to be indexed in pictures correspond to answers to the four questions: who?, what?, when?, and where?, and “each of these basic facets may then be subdivided into aspects based on Of in the specific sense, Of in the generic sense and About.” (Shatford 1986: 48) To illustrate let us take the following photograph:

The first level represents the literal general meaning that requires little or no cultural knowledge: it is a photograph of staircases and about a city street in winter. The second level of meaning represents the symbolic and specific, requiring some cultural background to discern: it is a photograph of St. Urbain street in Montreal, and about a disappearing winter beauty of iron craftsmanship.

Shatford’s who, what, when and where is the basis of a faceted classification of images, divided into answers to questions as to the Of and the About of a picture:

· “who or what, beings and objects, is this picture Of?”
· are these symbols, representations, abstractions or personifications, in other words, “what are they About?” (Shatford 1986: 50).

The Who Of is divided into the generic and specific. Berinstein gives the following illustrative example (86):

Who of Specifically	Who of Generally	Who About
Nancy Garman	Woman	Editor
Starship Enterprise	Spaceship	Exploring space, the future.

The “what” facet is broken down into Of (events and actions) and About (conditions and emotions). For example:

The above photograph is Of (general) a woman holding her hand, Of (Specific) Rossitsa Valkanova and About a film producer, support and perseverance.

The when facet includes specific linear and general cyclical time: “specific dates and periods (June 1885, Renaissance) and recurring time (Spring, Night)” (Shatford 1986: 53). The about aspect of when facet will rarely be used, it answers the question “Is the element of time represented in the picture a manifestation of an abstract idea? A symbol?” (Shatford 1986: 53). Berinstein interprets this as the “linking of ideas associated with time, such as ‘the end of the world’” (86) Similarly, the About aspect of the where facet will “only be present if the place symbolizes another place or an abstract idea (Heaven, Hell, Paradise)” (Berinstein 86). The where facet includes “geographic, cosmographic and architectural space” and is divided in the specific Of aspect that includes “locale, site, or place represented” and generic Of including “kinds of places, […] landscape, cityscape, interior, planet, jungle.” (Shatford 1986: 53)

Two Separate Approaches

Jörgensen identifies some broad research areas in the domain of image access closely related to each other: indexing and classification, image users and uses, image parsing. (Jörgensen 1999: 294) In the area of indexing and classification, there is a fundamental question of the appropriateness of textual indexing for a non-textual medium.

Studies of the research literature in image indexing and retrieval confirm that there are two distinct approaches to the problem: content-based and description-based. (Chu 1017) The content based approach is the automated extraction of textual and image properties whereas description-based refers to human indexing using captions and keywords for artist and image data. (Chu 1011). Content-based techniques are primarily the domain of computer scientists developing technologies for “direct retrieval of visual image content such as shape, texture, color, and spatial relationship” (Jörgensen 1999: 300) In content-based retrieval systems, instead of entering search terms, for example, the user selects an area of an image in order to narrow down the search and retrieve related items. (Jörgensen 1999: 300) Alternatively, the system fetches similar items based on quantifiable aspects of the image or its parts. Although content-based research is popular, it has not produced systems that satisfy end-user information needs. This is because properties of an image such as shape, texture, and color contribute to our understanding of an image, but do not define it. Text-based search techniques remain the most efficient and accurate methods for image retrieval (Jörgensen 1999: 302).