TThe paper considers the process of archive formation for the case of documents stored as images in information support systems for service maintenance processes concerning military weaponry and special equipment. This operation makes it possible to eliminate data redundancy in documents depending on the purpose of the system on the whole and characteristic markers of the images restored.

Рассмотрен процесс формирования архива документов, которые хранятся в виде изображений в системах информационной поддержки процессов сервисного обслуживания военного вооружения и специальной техники (ВВСТ). Данная операция позволяет устранять информационную избыточность документов в зависимости от целевого назначения системы в целом и характеристических показателей восстановленных изображений.

A concerted strategy of building up and running an organisation involved in service maintenance of military weaponry and special equipment (MWSE) is intended for maintaining their serviceable condition or good working order, with the specified performance indicators guaranteed, and providing the resources required for such maintenance. Lying at its base is a combination of functions performed by the operating organisation, repair agencies, and service centres. At the present stage this strategy is to be supported by:

In accordance with the strategy of electronic technology application, the main element of the system of service maintenance information support is an electronic remedial maintenance file. It is essentially a consolidated process document to be followed in the course of MWSE remedial maintenance. It contains the aggregate of accounting records and reporting instruments characterising technical state of the object of maintenance, required scope and composition of works performed, work performers, human and material resources required for recovery of serviceable condition (good working order) of the object of maintenance. These documents are accumulated in an electronic database throughout the effective period of contracts for MWSE service maintenance. Because all the documents of an electronic remedial maintenance file are scan copies of real accounting records and reporting instruments, which defy efficient compression using standard cataloguing tools, a task arises to develop special archiving facilities for such documents.

To decrease the volume of stored data, it is proposed to reduce information redundancy of such documents, which are essentially binary images, based on the use of binary frequency analysis.

A binary spectral analysis can be based on Walsh binary functions, which create an orthogonal basis with some or other method of ordering. In our case we shall use ordering by the frequency or the number of intersections of the values of zero axis basic function. Such basic functions can be written as follows:

Sw ={walw(i, j), i, j = 0, 1, ..., N-1}, (1)

where walw (i, j) - Walsh function;

w - index denoting ordering by frequency, as opposed to Hadamard (h) or Paley p ordering;

i - defines function number;

j - function argument [

N = 2n, n = 1, 2, 3, ....

In discrete representation, Walsh ordering is determined by matrix [

ri(u); υi - coefficients of binary representation of the numbers of lines and columns.

Based on the representations given, a binary spectral transformation, i. e. Walsh transform, can be built.

Walsh transform Hw for a certain block (n x n), where is defined by matrix equation

where Bx(n) - matrix of the coefficients of direct Walsh transform;

X(n) - initial image matrix.

A reverse Walsh transform is performed by the same matrix Hw( n) due to its orthogonal and symmetric property:

General structure of the process of compressing the scanned remedial maintenance file documents is given in Fig. 1.

Fig. 1. Image compressing process structure

When using each processing option, different transformation options are obtained, and hence, different results as per the measured parameters. For assessment of the considered images, the root mean square index of the initial and compressed image quality can be used. The mean squared error (MSE), applied for assessment of distortions of reconstructed images, is calculated by the formula [

where N - image dimensions in pixels;

M0 - element of the initial image grey-level array;

M - element of the reconstructed image grey-level array.

If the images are identical, then MSE = 0. When applying data compression, upon orthogonal Walsh transform completion, any insignificant elements, whose loss does not cause changes visible to human eye as compared with the initial image, are deleted (or discarded). The number of discarded elements is determined proceeding from matrix representation of the transformation spectrum and is assigned by a threshold, with the elements below it set equal to zero. The degree of deleted elements is calculated by the formula [

where C - total number of spectrum elements;

C0 - number of deleted elements.

This paper considers three technologies for reducing binary spectrum redundancy:

Fig. 2. Initial image (a) and its binary spectrum (b)

For the third action, standard archiver 7Zip was used in the experiments. An example of the initial image of document part and its binary spectrum is given in Fig. 2.

Table 1 contains binary spectrum analysis results (see Fig. 2). A reconstructed image obtained after quantisation of spectrum mid-frequency components is shown in Fig. 3. Such operation allows to transit from the 8-bit coding of spectrum coefficients to the 4-bit one, which contains sign of the number and the number of the level which its value is assigned to. From the analysis of data given in Table 1 and Fig. 3, it can be concluded that transition to the 4-bit coding has practically no effect on the reconstructed image quality.

Table 1

Obtained spectrum analysis

Q-ty of elements in range | Thresholds |
---|---|

-85...-75 (inclusive) = 3 | -77,5 |

-75...-65 (inclusive) = 11 | -67,5 |

-65...-55 (inclusive) = 61 | -57,5 |

-55...-45 (inclusive) = 174 | -47,5 |

-45...-35 (inclusive) = 434 | -37,5 |

-35...-25 (inclusive) = 1083 | -27,5 |

-25...-15 = 3672 | -17,5 |

-15...0 (inclusive) = 122 633 | Not used |

0...15 (inclusive) = 127 903 | Not used |

15...25 (inclusive) = 4106 | 17,5 |

25...35 (inclusive) = 1116 | 27,5 |

35...45 (inclusive) = 326 | 37,5 |

45...55 (inclusive) = 61 | 47,5 |

55...65 (inclusive) = 16 | 57,5 |

65...75 (inclusive) = 4 | 67,5 |

75...85 (inclusive) = 1 | 77,5 |

Fig. 3. Reconstructed image obtained after quantisation of spectrum mid-frequency components (MSE = 0.021476); high-frequency spectrum portion unchanged

Given in Fig. 4 are the results of deletion of insignificant spectrum elements according to threshold ±3 and the image reconstructed from that spectrum. Here, S = 56.43 %; MSE = 1.55. The number of discarded positive values is 73,960, negative - 73,971.

Fig. 4. The result of deletion of insignificant spectrum elements according to threshold±3 (а) and image reconstructed from this spectrum (S = 56.43 %; MSE = 1.55) (b)

Fig. 5 demonstrates the repeated experiment, as shown in Fig. 4, with threshold of ±6. Here, S = 80.24 %; СКО = 2.49. The number of discarded positive values is 105,369, negative - 103,122.

Fig. 5. The result of deletion of insignificant spectrum elements according to threshold±6 (а) and image reconstructed from this spectrum (S = 80.24 %; MSE = 2.49) (b)

From the data given it can be concluded that even under a high degree of image compression (over 80 %) the text readability remains satisfactory. Application of quantisation and spectrum components decimation allows to obtain the best result [

Table 2

Comparative characteristics of the obtained documents by volume

Description | Format | Size, byte | |
---|---|---|---|

before compression | in zip | ||

1. Initial image | .bmp | 263 222 | 106 320 |

Jpeg | 57 436 | 55 536 | |

2. Walsh transform spectrum (w/o compression and quantisation) | .bmp | 263 222 | 81 039 |

Jpeg | 165 590 | 165 605 | |

3. Walsh transform spectrum (after quantisation and compression) | .bmp | 263 222 | 15 738 |

JPeg | 99 021 | 96 319 |

It should be pointed out that after spectral processing the documents can be efficiently compressed with standard archivers. The compression degree at that can amount to approximately 30 times relative to the initial document.

Application of special document compression techniques allows to create compact archives of remedial maintenance files and promptly monitor the degree of execution MWSE service maintenance contracts.

Modelling in the MATLAB environment has proved good efficiency of the proposed methods and algorithms for compressing binary images that defy compression by conventional methods.

The authors declare that there are no conflicts of interest present.