LMMIP

IDENTIFYING MODELS OF SYSTEMS FOR IMAGE GENERATION USING A SMALL NUMBER
OF OBSERVATIONS
Main ideas: taking into account a special character of problems of estimating via a small number of observations, the algebraic theory of perturbations that requires no apriori probability distributions knowledge is applied and developed. Data informativeness is pre-analyzed. Additional non-statistic information is used to improve estimates.

Example: information technology of identyfing reconstructing filters using fragments of object images.

Major technological stages:

1.A fragment with desired (stepped) infensity function is extracted in the image of a known object.
2. Fragment is reconstructed (computer retouching) from its a priori definition.
3. Reconstructihg filter is identified using the original and reconstructed fragments by the developed methods.
4. The entire distorted image (or a series of images) is reconstructed, given unchanged registration conditions .

Distorted image of the known object with extracted fragment


Result of reconstruction of the extracted fragment (magnified)

Original distorted image	Reconstructed image

Publications:
1. V. V. Sergeyev, V. A. Fursov, S. I. Parfenov, Proceeding of the IV-th ROAI-98 conference. Novosibirsk, P. I, 1998. P. 378-381.
2. V. A. Soifer, V. V. Kotlyar, V. A. Fursov, Third conference of the Volga-region RARAS Center, Sarov, 1998. P. 108-109.
3. Journal: Proc. SPIE, 1997, v. 3087, pp. 34-44.
4. Pattern recognition and image analysis, 1998, v. 8, N2, pp. 269-271.

LMMIP