Modifying Graphics Images

Abstract

This paper reports on the development of a text understanding program for a minicomputer with 8K memory to facilitate modifying graphics images. Two versions of an interactive language are described. An early system used a non-linguistic method for picture variation. Later emphasis was placed on conversational linguistic structure. The early language used regular decomposition to approximately locate standard display templates. The later language allowed conversational dialogues between the user and image modification routines. We present an implementation of it that retains context during a dialogue and enables relational adjustments of facial features. Imprecise feature judgements issued by the user are used to modify images. Ambiguities encountered by the program are resolved by interrogating short term memory buffers and hash table data. Language and hardware features are combined by data structures that interface display processor instructions and requests generated by the interpreted text. Updating these structures is discussed in terms of relationships between structure elements.