Application of the SP theory of intelligence to the understanding of natural vision and the development of computer vision

摘要

The SP theory of intelligence aims to simplify and integrate concepts incomputing and cognition, with information compression as a unifying theme. Thisarticle discusses how it may be applied to the understanding of natural visionand the development of computer vision. The theory, which is described quitefully elsewhere, is described here in outline but with enough detail to ensurethat the rest of the article makes sense. Low level perceptual features such as edges or corners may be identified bythe extraction of redundancy in uniform areas in a manner that is comparablewith the run-length encoding technique for information compression. The concept of multiple alignment in the SP theory may be applied to therecognition of objects, and to scene analysis, with a hierarchy of parts andsub-parts, and at multiple levels of abstraction. The theory has potential for the unsupervised learning of visual objects andclasses of objects, and suggests how coherent concepts may be derived fromfragments. As in natural vision, both recognition and learning in the SP system isrobust in the face of errors of omission, commission and substitution. The theory suggests how, via vision, we may piece together a knowledge of thethree-dimensional structure of objects and of our environment, it provides anaccount of how we may see things that are not objectively present in an image,and how we recognise something despite variations in the size of its retinalimage. And it has things to say about the phenomena of lightness constancy andcolour constancy, the role of context in recognition, and ambiguities in visualperception. A strength of the SP theory is that it provides for the integration of visionwith other sensory modalities and with other aspects of intelligence.