72 PSEUDOGRADIENT OPTIMIZATION IN THE PROBLEM OF IMAGE INTERFRAME GEOMETRICAL DEFORMATIONS ESTIMATION1 G. V. Dikarina2, G. L. Minkina2, A. I. Repin2, A. G. Tashlinskii2 1 Ul’yanovsk State Technical University 432027, Russia, Ul’yanovsk, ul. Severnyi venets, 32, phone (88422)430974, e-mail: tag@ulstu.ru When pseudogradient estimating of image parameters the estimate convergence character and computational expenses depend on the image samples local sample size to be is used for the pseudogradient finding. In the work the solving local sample size optimization problem on criterion of computational expenses minimum when image interframe geometrical deformations estimating is considered. When solving image interframe geometrical deformations (IIGD) estimation problem pseudogradient procedures (PGPs) are applied [1] ˆ t 1 ˆ t Λt 1t 1 Q Zt 1, ˆt , where – vector of IIGD parameters to be estimated Ζ (1) {z (j1) } and Ζ( 2) {z (j2) } ; – pseudogradient of the goal function Q characterizing estimation quality; Λ t – gain matrix determining a parameters estimate increment at the t -th iteration; Z t 1 – local sample of images Ζ (1) and Ζ( 2) samples to be used to find at the t 1 -th iteration; Ζ (1) and Ζ( 2) – images to be observed. It is obvious the local sample size (LSS) determines computational expenses for PGP realization in many respects. Let us consider the problem of PGP LSS optimization on criterion of computational expenses minimum when estimating of one parameter of IIGD. At that the module of the error ˆ of parameter estimate in the estimation process have to vary from some maximum max to some minimum min Let us denote computational expenses with d t which are necessary to perform the PGP t -th iteration when the local sample size is equal to t and total computational expenses which are required for decreasing of the error from max to min with DT d t , T t 1 where T – number of iterations which is required for condition T min fulfillment. Besides that let us denote the relation of computational expenses at the t -th iteration to the mathematical expectation Mv ̂ of the parameter estimate convergence rate (computational expenses per one Mv ̂ of the parameter estimate convergence rate) with reduced computational expenses d* t d t , Mv ˆ where Mv ˆ ˆ wt 1 ˆ wt ˆ dˆ numerically equal to the difference between mathematical expectations of estimate at the t -th and t 1 -th iterations; wt 1 ˆ – values, where – an exact parameter value. _______________________________________________________________________ 1 is 0 The work was supported by the Russian Foundation for Basic Research, project no. 07-01-00138-a 73 probability distribution density (PDD) of the estimate ̂ . Minimum of computational expenses at each iteration will be ensured at the LSS enabling minimum of reduced computational expenses. Such LSS for the t -th iteration will be assumed optimal and denoted with *t : *t k d* k min d* t , t 1, 2, ..., k , ... . (1) The parameter estimate error for T iterations of PGP has to change from max to min , so the choice of LSS at each iteration according to (1) ensures minimum of the total computational expenses Dmin T d *t . T t 1 Let us consider the algorithm of finding optimal dependence LSS on the number of iterations (2) *t , t 1, T . For concreteness let us accept some assumptions. Let us part the computational expenses d t at one iteration performing by the algorithm on two constituent: expenses to form local sample ( d L t ) and other computational expenses ( d A t ) d t d L t d A t . Besides for simplicity let us assume that expenses d L t for forming local sample are proportionate to size t of the local sample: d L t t d , where d – computational expenses at 1 . Then, d d t d A t t . d Assuming mentioned limits for algorithm construction to find (2) it is necessary to find the expression for parameter estimate convergence rate mathematical expectation Mv ̂ calculation. It can be done through PDD of the error . However in this work let us consider the simplified method of Mv ̂ calculation based on the usage of step t of parameter estimate variation at the t -th iteration, mathematical expectation M t 1 of the estimate error and probability of estimate improvement ( t 1 ) and estimate deterioration ( t 1 ) at the t 1 -th iteration [2]. Really the forecast of the mathematical expectation M t of estimate error at the t -th iteration can be represented in the following form Mt Mt 1 t t 1 t 1 . Then Mvˆ Mt 1 Mt t t 1 t 1 . Thus the mathematical expectation Mv ̂ of parameter estimate convergence rate at the t th iteration can be expressed through t 1 , t 1 and step t . Let us assume 0 t 1 0 , then Mv ˆ t 2 t 1 1 . t t 1 , t 1 2 t 1 1 – estimate where improvement coefficient (EIC). In the beginning of calculation as the estimate initial approximation ( t 0 ) the maximum error of parameter 0 max is specified. Then for the first iteration simulation a variable t corresponding to the iteration number increases by one. When simulating any iteration the initial value of the LSS is equal to one 1 . For the given values и t the EIC t , , computational expenses d and estimate convergence rate Mv at the given error t and LSS are calculated. Then reduced computational expenses d* which are compared with expenses obtained with the LSS d* are found. If d* d* 1 then LSS increases by one and the next value d* 1 is calculated. If d* d* 1 then optimal value of LSS for the t -th iteration, equal to *t 1 remember. Then a new estimate error is calculated t 1 t Mt . If t 1 min , optimal LSS for the next t 1 -th iteration is found, if 74 t 1 min then the algorithm operation has been finished. Examples of optimal LSS as function on error h of two images parallel shift estimate at the parameter value d A t d equal to 25 and 16,6 are given in Fig. 1,a and Fig. 1,b correspondingly. When calculating the adaptive model of images z (j1) s (j1) (j1) , z s ( 2) j (1) j to be observed was s – desired random field () (j2) accepted, where (j1) with known monotonically decreasing autocorrelation function; (j1) , (j2) – interfering independent Gaussian random 2 fields with zero mean and equal variances . In the given plots the lower curve corresponds to the noise absence, middle – signal-noise relation 8, upper – 15. 25 Analysis of dependences shows that in the absence of noise optimal LSS monotonically decreases when decreasing estimate error. In the presence of noise with small errors LSS increases again that is due to EIC decreasing. Besides at the smaller part of computational expenses the range of optimal LSS change to form LSS is larger. In the table numerical results showing the loss in computational expenses at the constant LSS equal to 1, m and m 3 in comparison with the case of optimal LSS usage are presented, 1 T * where m int t – average LSS at the T t 1 iteration, T – total number of PGP iterations. 0, 2, 5, 10, 20. It is obvious when increasing signal-noise relation the loss decreases. Table. The loss in computational expenses in comparison with the case of optimal LSS usage, per cent Relation signalnoise * g 0 g2 g 5 g 10 g 20 20 15 10 t 5 1 31 61 91 121 151 а) * Local sample size to be used 1 m 3 m m 3 48,8 4,63 3,78 5,75 60,9 0,58 0,43 0,81 59,8 1,23 1,3 1,91 58,2 2,18 1,74 3,22 56,7 2,72 2,53 4,46 Thus the proposed method enables to solve problem of optimal LSS calculation on criterion of computational expenses minimum. References 16 12 8 t 4 1 31 61 91 121 151 b) Fig. 1. The dependence of optimal LSS on iteration number 1. Tashlinskii Alexandr. Computational Expenditure Reduction in Pseudo-Gradient Image Parameter Estimation / Computational Scince – ICCS 2003. Vol. 2658. Proceeding, Part II. – Berlin: Springer, 2003. – Pp. 456-462. 2. Tashlinskii A.G. The Efficiency of Pseudogradient Procedures for the Estimation of Image Parameters with a Finite Number of Iterations / Pattern Recognition and Image Analysis, Vol.8, 1998. – Pp. 260-261.