Résumé de séminaire

Résumé :

Weighted finite automata (WFA) compute real functions over words of the input alphabet. When the words are interpreted as addresses of pixels, the functions represent digital images. In this talk we talk about this image representation, giving several examples. We explain how the minimum state WFA can be found for a given image. A practical image compression algorithm based on these ideas is described. Weighted finite transducers (WFT) are finite state devices to perform certain image transformations. We show that several useful image operations can be implemented in terms of WFT including zooming, cropping, FIR filtering, etc. The implementation of image operations using WFT has the advantage that the operation can be applied to any WFA represented image directly in the WFA form, without decoding the image first. The result of the operation is then precise and it can be correctly decoded at arbitrarily high resolutions.

Références :

Le professeur J. Kari est affilié à : Dept. of Mathematics, Univ. of Turku, Finland, et Dept. of Comp. Sci., Univ. of Iowa, USA.

Home page : http://www.cs.uiowa.edu/~jjkari/