RU2350021C1 - Communication path muting mode for 2-d image signalling - Google Patents

Abstract

FIELD: physics.

SUBSTANCE: pickup signal is coded while processed through direct Hadamard transform ordered in compliance with Gray transformation. Herewith direct Hadamard transform coefficient matrix is sequentially processed with X-sweep operator L and complementary operator L* representing transposed matrix L. With using backward X-sweep operators L^- and L*^- respectively, decoding is performed. It is followed with summation of two formed Hadamard transform coefficient matrixes of original image and inverse Hadamard transform of total matrix. Method improves muting efficiency due to summation of two realisations of spectral noise distribution images with mutually orthogonal coordinates in image field. Therefore total dispersion of two realisations of noise sequences acquires synergetic effect ensured with nonlinear dispersion distribution across image field and sharp increase in signal/noise ratio within small dispersions.

EFFECT: higher communication path muting efficiency for 2D image signalling.

3 dwg

Description

The invention relates to the field of information transmission, specifically to methods for transmitting signals in optical communication channels, and may find application in creating promising formats for transmitting information through optical channels.

Actual tasks in this area are the possibilities of increasing compression, noise immunity and speed during signal transmission. The problem of filtering various noises (interference) is usually solved by using frequency filters, which leads to distortions in the bandwidth of the communication channel and, thereby, to the violation of the completeness of the reconstructed image. However, at present, for noise suppression, methods are often applied that include integrated image processing with transmission of spectral coefficients corresponding to various systems of basis functions used for transforming the original images over the communication channel, for example, methods using integral transforms such as the Walsh-Hadamard transform.

A known method of signal transmission using the Walsh-Hadamard transform (US application No. 20060067413 from 03.30.2006, H04K 1/10, H04L 27/06), in which the Walsh-Hadamard transforms are used to encode signals in such a way that the dimension and type of the matrix transformations are affected by the Sherman-Morrison algorithm in accordance with the covariance noise matrix. The use of the Hadamard transform in this method for encoding signals during their transmission over the communication channel with transformation of the dimension and type of the transformation matrix using the Sherman-Morrison algorithm in accordance with the covariance noise matrix allows minimizing the mean square error.

This method assumes the existence of a multi-channel network with the possibility of multiple access with code division multiplexing, which significantly complicates and increases the cost of the information transmission system. In addition, from the point of view of noise immunity, this method relies on traditional ideas about the influence of the signal-to-noise ratio on the quality of the output signal, under which it is assumed that the amount of minimized noise (its dispersion) is the same for the entire array of transmitted signals, regardless of the information content of the image field , which reduces the ability to maximize the signal-to-noise ratio in the most informative areas of the transmitted image.

The prototype of the proposed method is a method of spatial filtering of temporary noise (B. Podlaskin. Spatial filtering of temporary noise when implementing the Hadamard transform on a photodetector. ZhTF, v.77, v.4, 2007, p.139), including integrated image processing with decomposition them according to the system of Walsh-Hadamard basis functions of dimension N ordered in accordance with the Gray transform, transmitting the obtained spectral coefficients through the communication channel in the presence of additive noise, performing the inverse transform Hadamard with restoration of the original image F and localization of noise in different parts of the screen. The authors found that changing the organization of the system of basis functions by rearranging the rows in the direct Hadamard transform matrix H (i.e., changing the order of transmission of the spectral components) affects the distribution of the noise spectrum in the image space. In this case, it is possible to concentrate the maximum noise distortion of the image in different areas of the screen. In other words, it is possible to perform spatial localization of the noise dispersion by concentrating the bulk of the noise in the least informative parts of the reconstructed image. To transmit signals over the communication channel, this method uses the direct horizontal scanning operator L, which converts the matrix of direct Hadamard transform coefficients C, obtained on the transmitting side, into a one-dimensional vector W with a dimension equal to the square of the dimension of the direct Hadamard transform. When passing through the communication channel, additive noise ψ is superimposed on the signal W, as a result of which the signal W 'is recorded at the receiving side. The one-dimensional vector W 'with the help of the inverse horizontal scanning operator L ^{- is} transformed into the matrix of direct Hadamard transform coefficients C', over which the inverse Hadamard transform H ^- is produced, as a result of which the image F 'representing the restored original image F and the image is displayed on the receiving screen noise, in the form of its spectrum in the basis of Walsh functions, localized in certain areas of the image field. Thus, the processing and transmission of the optical image is described by the following expressions:

Fh = c

CL = W

W + ψ = W '

W'L ^- = C '

C'H ^- = F '= F + H ^- ψL ^- ,

or

H ^- ((FH) L + ψ) L ^- = H ^- FH + H ^- ψL ^- = F '= F + H ^- ψL ^-

Since the operator L is the horizontal scanning operator, the spatial localization of the channel temporal noise in accordance with the line-by-line arrangement of the sequentially ordered basic Walsh functions is of a pronounced horizontal character, i.e. the maximum values of the noise variance are located along the lines corresponding to the position of the Walsh functions with minimal sequents. The method allows to receive a signal-free region of the screen space located in the center of the screen at the top or bottom of the screen when receiving a signal, i.e. produce spatial lowercase noise localization.

The disadvantage of the prototype is the implementation of spatial localization of noise in only one dimension, which does not make it possible to obtain a noise-free region of the screen space of a significantly larger size with comparable values of the width of the spatial spectral noise characteristic and the dimension of the transformation.

The invention solves the problem of increasing the noise suppression efficiency of a communication channel when transmitting a two-dimensional image signal.

The problem is solved by a method of suppressing the noise of a communication channel when transmitting a two-dimensional image signal, including encoding the original signal by acting on it with the direct Hadamard transform ordered in accordance with the Gray transform, sequentially affecting the matrix of direct Hadamard transform coefficients by the horizontal scanning operator L and the additional operator L *, representing a transposed matrix L, for the formation of one-dimensional vectors W and W *, sequential transmission of coefficients of these vectors over the communication channel, decoding the output signal obtained in the form of coefficients of the total vector (W + W * '), in which the total vector is divided into components W' and W * ', transform them using the inverse horizontal scanning operators L ^- and L * ^- , respectively, summarize the two formed matrices of Hadamard transform coefficients of the original image, after which the inverse Hadamard transform is performed on the total matrix.

Developing the idea developed by the authors of the present invention and embodied in the prototype method, the authors found that the degree of spatial localization of temporary noise of the communication channel (noise suppression efficiency) can be significantly increased by summing in the image field two realizations of images of the noise spectral distribution, the coordinates of which are orthogonal to each other, as a result of which the dispersion summation of two realizations of noise sequences acquires a supertotal effect due to nonlinear distribution of dispersion over the image field and a sharp increase in the signal-to-noise ratio in the region of small dispersion values.

In this case, the expressions describing the processing and transmission of the image take the form:

Fh = c

CL + CL * = W + W *

(W + ψ) + (W * + ψ) = W '+ W *'

W'L ^- + W * 'L * ^- = C' + C * '

С'Н ^- + С * 'Н ^- = F' + F * '= 2F + H ^- ψL ^- + H ^- ψL * ^- ,

or:

H ^- ((FH) L + ψ) L ^- + H ^- ((FH) L * + ψ) L * ^- = H ^- FH + H ^- ψL ^- + H ^- FH + H ^- ψL * ^- = F '+ F * '= F + H ^- ψL ^- + F + H ^- ψL * ^- = 2F + H ^- ψL ^- + H ^- ψL * ^- ,

where F is an array of data (matrix) intended for transmission over a communication channel (original image);

H is the direct Hadamard transform matrix;

C is the matrix of coefficients of the direct Hadamard transform;

L - line scan operator;

L * is an additional operator, which is the transposed matrix of the operator L;

W is the one-dimensional vector obtained by the operator L transforming the matrix of coefficients of the direct Hadamard transform C on the transmitting side;

W * is the one-dimensional vector obtained by the operator L * transforming the matrix of coefficients of the direct Hadamard transform C on the transmitting side;

ψ is the additive noise;

W 'is the one-dimensional vector corresponding to W registered at the receiving side;

W * 'is the one-dimensional vector corresponding to W * recorded at the receiving side;

L ^- is the operator that converts the one-dimensional vector W of dimension N ² into a matrix C 'of dimension N;

L * ^- is the operator that converts the one-dimensional vector W * 'of dimension N ² into a matrix C *' of dimension N;

C 'is the matrix of coefficients of the direct Hadamard transform obtained by acting on W inverse line scan operator L ^- ;

С * 'is the matrix of coefficients of the direct Hadamard transform obtained by acting on the inverse horizontal scan operator L * ^- on W *';

F 'is the image representing the restored original image F and the noise image ψ when exposed to the operator L ^- ;

F * '- the image representing the restored original image F and the noise image ψ when exposed to the operator L * ^- .

Involving an additional operator L *, which is the transposed matrix of the operator L, reading the coefficients of the direct Hadamard transform H over the image F and forming the one-dimensional vector W *, as well as involving the additional operator L * ^- , which is a transposed matrix, in the encoding operation of the image operator L ^-, converting the one-dimensional vector W * 'dimension N ² matrix C *' dimension N, leads to the fact that the reduction of the image pho ma useful signal is stored, and its capacity is doubled (2F) through the use of double-communication session duration (because in this case there is a transfer of two vectors W and W * instead of one). Since the additive noise communication channel distorts the signals corresponding to the transmitted coefficients Hadamard affects the vectors W and W * is statistically independent, and the covariance matrix of the noise variance because of the orthogonality of vectors L ^- and L * ^- becomes a two-dimensional dispersion distribution reduced noise of the image field becomes substantially non-linear so that the concentration of maximum dispersion can be concentrated on uninformative areas of the image, for example in its corners. As a result, at a given threshold of minimum dispersion, determined by the minimum acceptable value of the signal-to-noise ratio, the noise suppression efficiency is significantly increased by its spatial localization. Assessing the effectiveness of noise suppression by spatial localization is characterized by the diameter of a circle inscribed in the image field, the dispersion value in which does not exceed a predetermined threshold of minimum dispersion.

The method is as follows.

At the first stage, a two-dimensional image F is read in the form of a matrix of input parameters with a dimension of N × N. The matrix F is encoded - first, by direct Hadamard transform H, using a Hadamard matrix with dimension N × N, which is ordered in accordance with the Gray transform:

.

.

, с размерностью N×N, содержащая коэффициенты прямого преобразования Адамара над матрицей F, представляющая собой массив данных, предназначенных для передачи по каналу связи (HF=C). Далее производят преобразование матрицы С в линейную последовательность данных (W+W*) с помощью последовательного воздействия оператора L=|a_k|, где k=j+N(i-1), с размерностью N² и оператора L*=|b_k|, где k=i+N(j-1), с размерностью N², т.е. CL+CL*=W+W*. Затем производят последовательную передачу элементов суммарной последовательности (W+W*) с размерностью, равной 2N², по каналу связи в виде электрических сигналов, подаваемых на входное устройство канала связи и согласованных по форме и частоте следования с параметрами канала, обладающего собственным шумом ψ, аддитивным по отношению к передаваемому сигналу. На приемной стороне канала производится регистрация выходного сигнала (W'+W*'), включающего входную последовательность (W+W*) и аддитивный шум:As a result of this Hadamard transform, a matrix of direct transform coefficients is created

, with the dimension N × N, containing the coefficients of the direct Hadamard transform over the matrix F, which is an array of data intended for transmission over the communication channel (HF = C). Next, the matrix C is transformed into a linear data sequence (W + W *) using the sequential action of the operator L = | a _k |, where k = j + N (i-1), with dimension N ² and the operator L * = | b _k |, where k = i + N (j-1), with dimension N ² , i.e. CL + CL * = W + W *. Then, sequential transmission of the elements of the total sequence (W + W *) with a dimension equal to 2N ² is made over the communication channel in the form of electrical signals supplied to the input device of the communication channel and matched in form and frequency with the parameters of the channel having its own noise ψ, additive with respect to the transmitted signal. On the receiving side of the channel, the output signal (W '+ W *') is recorded, including the input sequence (W + W *) and additive noise:

W '+ W *' = (W + ψ) + (W * + ψ). At the receiving side, the output signal is decoded. At the first stage, the received signal (W + ψ) + (W * + ψ), which has a dimension of 2N ² , is divided into two successive components in time (W + ψ) and (W * + ψ), after which they are transformed with using the inverse operators L ^- and L * ^- , respectively, so that two matrices with dimensions N, C 'and C *' are formed, formed from the Hadamard transform coefficients transmitted over the channel: W'L ^- + W * 'L * ^- = C '+ C *'. The inverse Hadamard transform is performed over the sum of matrices (C '+ C *'): C'H ^- + C * ^- H ^- = F '+ F *' = 2F + H ^- ψL ^- + H ^- ψL * ^- , as a result of which the original image and the image of the spectrum of the additive noise of the channel are restored in the basis of the Hadamard transform with non-linear dispersion distribution over the image field and its concentration in the least informative parts of the reconstructed image, which leads to suppression of the noise of the communication channel when transmitting two-dimensional image signals. An example of a specific implementation.

The method was carried out by transmitting using a fiber optic communication channel a signal of the form:

Using the sequential action of the operator L, which is a sequential reading of the matrix elements along its rows, and the operator L *, which is a sequential reading of the matrix elements along its columns, the matrix C was transformed into a linear data sequence (W + W *):

This sequence in the form of electrical signals was applied to a modulator of a fiber-optic communication channel that converts electrical signals into optical ones. As a result, the elements of the total sequence (W + W *) were sequentially transmitted over the communication channel in the form of optical signals that were matched in form and frequency with the parameters of the channel with its own noise ψ, additive with respect to the transmitted signal. The implementation of noise ψ with a Gaussian distribution and dispersion of 0.5 was as follows:

On the receiving side of the channel, an optical detector (photodiode) recorded the output signal (W '+ W *'), including the input sequence (W + W *) and additive noise ψ, W '+ W *' = (W + ψ) + ( W * + ψ):

using a computer, the Hadamard transform was inverse, which restored the original image and the spectrum of the additive noise channel in the basis of the Hadamard transform, with a nonlinear dispersion distribution over the image field and its concentration in the least informative parts of the reconstructed image, which suppressed the noise of the communication channel when two-dimensional image signal transmission F:

The results of applying the method are graphically presented in FIGS. 1, 2, 3.

Figure 1 shows the screen image of the original two-dimensional optical signal transmitted over the communication channel. The signal is distributed over the central part of the screen, the color intensity corresponds to its power.

Figure 2 shows the image of the output signal obtained by using the method described in the prototype. It can be seen that, in addition to the signal in the central part of the screen, noise is distributed along the lines in the upper and lower parts of the screen, and it practically closes with the signal, which leads to signal distortion.

Figure 3 shows the image of the output signal obtained by using the proposed method. One can clearly see a significant decrease in the area occupied by noise (in the corners of the screen), as well as a decrease in its intensity (light color).

Thus, from a comparison of FIGS. 2 and 3, it is obvious that the degree of noise localization and the signal-to-noise ratio in the region of the transmitted signal, i.e. the noise reduction efficiency when using the proposed method is significantly higher than when using the prototype method.

Claims (1)

A method of suppressing the noise of a communication channel when transmitting a two-dimensional image signal, including encoding the original signal by acting on it with the direct Hadamard transform ordered in accordance with the Gray transform, sequentially affecting the matrix of direct Hadamard transform coefficients with the horizontal scanning operator L and the additional operator L *, which is the transposed matrix L, for the formation of one-dimensional vectors W and W *, sequential transmission of the coefficients of these vectors in k communication channel, decoding the output signal obtained in the form of coefficients of the total vector (W '+ W *'), in which the total vector is divided into components W 'and W *', transform them using the inverse horizontal scanning operators L ^- and L * ^- , respectively, the two formed matrices of Hadamard transform coefficients of the original image are summed, after which the inverse Hadamard transform is performed on the total matrix.

Images