Zhisnyakov A.L., Fomin A.A., Privezencev D.G., Baranov A.A.

Murom Institute of Vladimir State Univercity

STRUCTURE ANALYSIS OF IMAGES SEQUENCES

The work purpose is approach construction to the description of the digital images, based on analysis of sign group behavior on images sequence.

Image is a function f(x,y), which is determined on subset P of plane R2 and takes on a real values.

Let's consider we can compare each image f  to finite signs set X =  {X1, X2, …Xn}, that unambiguously determines f into multiple set of another images given on P.

We name some ordered images sequence as {fn} and set index for each element {fn}={f0, f1,… fn}. (1)

We limit all multiformity of existing sequences and view case that we can describe sequence forming by some operator. Thus every image from sequence is a result of influencing of operator T on the previous one.

As each image from initial set f is unambiguously characterized by signs set X, we can construct the set {Xn} = {X0, X1, X2 …} identical with (1).

Let's consider if there is some images sequence {fn}, that is ordered by any signs set X, then adjoining images can have some similarity degree. In other words if any sign presents on the image fk Î{fn}, probably it will presents on the neighbouring images fk-1 Î{fn} è fk+1 Î{fn}, k = 1..N-2. Naturally the images are more closely among themselves, the longer there is the images sequence

{… fk-2, fk-1, fk, fk+1, fk+2...},    k = 1..N-2,                                        (2)

which contains same sign.

Therefore we can talk about sign inheriting factor on images sequence.

At the same time it is obvious, as relationship between images in the sequence will decrease depending on distance between ones (increasing image index difference), some signs, that distinctly show on the one image, can be less visibly on the another images or vanish quite.

It is characterized as the signs mutability on the images sequence.

Let's consider sequence consisted of elements {Xn} that define same  sign (in example sign with index i) for viewing all images from initial sequence

{x(i)n} = {x0(i), x1(i),… xn(i)}.

We named operator that defines sign x(i) entry level into fuzzy set X(i) as sign of attribute operator. It represents conformity degree of single sign realization to the limit (etalon) value on the images sequence.

mi : mi[xj(i)]®[0,1].

Then we can hold up as a conformity to each signs set Xj of image fjÎ{fn} the vector  mÕj = (m1õ1, m2õ2, .. mkõk )Ò where k is signs amount.

For every sign we can introduce the threshold operator

Ã: Ãmi[xj(i)]®{0,1}.                                                                       (3)

As a result of application of (3) to initial vector Xj that characterized image fi, we get some vector Wj consists of ones and zeroes:

W = (w1, w2,… wê)Ò, wi Î {0,1}, i = 1, 2,…k.

Let we have two neighboring images fm and fm+1 from sequence {fn}.  Vectors Wm and Wm+1 agree with them.

Let’s consider pairs of elements (wmi, wm+1i), i = 1, 2,..k. Elements coincidence in such pairs is there are same sign presented on both images or it is not presented on ones.

Sequence mutability is the process of old signs losing or new signs acquiring while transiting to each next sequence image.

To get mutability characteristic we can use the expression

,

where Å is inequivalence operation (exclusive disjunction).

Inheriting of images sequence is signs holding process while transiting to the next image of sequence.

Taking into consideration relation between mutability and inheriting terms we can introduce sign mutability characteristic in this view:

.

Sign x(i) distribution curve on the images sequence {fn} is graph of function mi(j) º mi[xj(i)] that takes on a discrete values on the points jÎ{0,1,..n}.

To realization sequence of some sign on the images sequence the sequence consists of ones and zeros agrees

Q = {q1, q2,… qn}, qj Î {0,1}, j = 1, 2,…n.                                   (4)

Index i of sequence Q element for which condition (qi-1 = 0) Ù (qi = 1) is true we can name sign appearance point.

Index j of sequence Q element for which condition (qj-1 = 1) Ù (qj = 0) is true we can name sign disappearance point.

The result of operator à application to sequence of function mi(j) values we can name normed curve of sign x(i) distribution on the images sequence {fn}.

The part length of normed curve of sign x(i) distribution on which the function Ãmi(j) takes on value equal one we can name depth of sign x(i) enclosure on the images sequence {fn}.

Hence we can think image f  can be characterized by both signs X set and mutability features of this signs on the images sequence than derived from successive multiple application of operator T to the analyzed image. At that one sign that is congruous for two different images will behave on their sequences differently, because of different mutability and inheritance character that is defined by images features.

References

1. Zhiznjakov A.L., Sadykov S.S.,Gai V.E. Analysis of impact behavior of feature’s group at image sequence. 9th International Conference “Pattern Recognition and Image Analysis: New Information Technologies” (PRIA-9-2008): Conference Proceedings. Vol.2.-Nizhni Novgorod, 2008.-404 p.(pp.365-366)