APPROACH TO ELABORATION OF IMAGES PROCESSING SOFTWARE
[1. Інформаційні системи і технології]
Автор: Kovivchak Yaroslav, Cand. Tech. Sc., Assoc. Professor, Department of Automated control systems, Assoc. Professor, National University «Lviv Polytechnic», Lviv; Dubuk Vasyl, Cand. Tech. Sc., Assoc. Professor, Department of Automated control systems, Assoc. Professor, National University «Lviv Polytechnic», Lviv; Mishak Roksolana, Department of Automated control systems, Student, National University «Lviv Polytechnic», Lviv
In many practical cases connected with transmission and saving of graphical information the necessity of processing the image files arises even not for different images but for group of similar images also.
One of the most important forms of files with images processing is the compression. For such operation the solution may be found by means of cutting of information surplus concerning separate images and their groups.
With the development of information technologies, technical progress in the sphere of microelectronics, optics, telecommunication technologies the new opportunities for reception, processing and transmission of high quality images are opened. As for today, in practice many different devices for receiving, processing and transmission of images with high discriminability are used in space vehicles for remote sensing of Earth, in airplanes, radars, hydrosonars, in television etc. So, the problem of development of effective tools for reduction of sizes of image files with minimum possible loss of quality is still actual.
Methods of compression of bitmapped images may be devided conditionally into two large groups : compression with losses and without losses. Methods of compression of images without losses provide low coefficient of compression, but foresee the recreation of initial values of pixels of initial image. Methods of compression of images with losses provide high coefficients of compression, but does not give an opportunity to recreate initial image with precision to the pixel.
The majority of known methods of image compressions provides for division of image into separate blocks (clusters) and compression of resulted blocks by means of corresponding algorithms.
So, the perspective and actual is not only the development of algorithms of image processing, but improvement of methods of optimal images clustering.
In practice for image compression different methods of clusterisation are used. As already known, methods of clusterisation are devided into hierarchical and unhierarchical. The result of application of hierarchical algorithms is development of clusters tree, root of such is all selection and leaves are different clusters.
The algorithms, based on the search of optimal division of sets of objects to the clusters (groups) became very popular for solving the problem of clusterization. Such algorithms form basis of unhierarchical methods of clusterization. Depending on initial task corresponding methods of clusterization are used.
One of the set of main problems of cluster analysis is the determination of opti-mal quantity of clusters. At many algorithms of unhierarchical clusterization this parameter is an input value.
It is necessary to underline that results of division of initial selection onto clus-ters may be differ substantially inter se depending on selected quantity of clusters and methods of clusterization. So, the selection of quantity of clusters at a division of image will influence considerably onto quality of resulting image and the file size.
The approach to elaboration of software tool for processing of bitmap images by means of their clusterization with algorithm K-means [2, 3, 4] using is presented in this work. This method is widely spreaded and the most invesigated among other methods of clusterization. The popularity of K-means method based on its main advantages: simplicity, flexibility, rapid convergence.
K-means use algorithm of image compression with losses, so to renew the initial image after it processing is impossible. So, the more coefficient of compression lead to the more difference between initial and processed images. Futhermore, the users can independently select the coefficient of compression for each other image.
The idea of K-means algorithm for image clusterization is based on finding of k-cenered pixels at image. So, every pixel of image in set of data will belong to some k-set with minimal Euclides distance. Every pixel is represented by 3 bytes (RGB) in coloured pictures. In this case the value of possible color of pixel may be between borders from 0 to 255. The total quantity of colors, which the pixel of image can represent accordingly is 16 777 216. A human eye at perception of color can not to distinguish such quantity of colors. So, using of this feature of visual perception by human eye of color images by means of algorithm of clusterization K-means the base set of values of colors of pixels for every different image may be distinguished. The near values of colors of pixel at image are represented by values of colors from formed set of values. On Fig.1 the conceptual model of software tool operation is presented.
Fig. 1. Conceptual model of software tool operation.
The use-case diagram and diagram of activity during projection and develop-ment of software tool for image compression based on clusterization are developed. The use-case diagram represents the main cases (functions) of system and sequence of their implementation with processing of users queries.
Fig. 2. Use-case diagram of software tool.
The Fig. 2 presents the use-case diagram of software tool for compression of image. The subject is the user, which cooperates with the next set of main cases: processing of image, selection, clustering, compressing on basis of clusterization with initial determined quantity of clusters of color, sending.
For image processing with compression one must select image, input number of clusters for quanting of color. After that the compression of size of image file by means of clusterization with algorithm K-means is executed.
The diagram of components for developed software tool was elaborated also and presented on Fig. 3. It represents the architecture of the system, so to provide the connection between main software components of the system.
Fig. 3. Diagram of components of software tool.
The elaborated software tool includes the main basic components: component of user interface; component of selection of image; component of image, which must be processed; component of input of clusters quantity; component of clusterization; component of compressed image; component of saving of compressed image. As the results of the elaboration the components of the software were realized.
On Fig. 4 the main window of graphical user interface of software tool for image compression on the basis of clusterization is presented. By means of main window of interface of software tool user may select the initial image for file size compression in which it will be saved after compression. The quantity of clusters, which will be pixels-centroids for corresponding image must be inputed into main window also. After that the compression of image will be fullfilled.
The processed compressed image may be looked over and saved at selected memory device. The users of software tool may select the necessary quality of compressed image for every different case. As the results of investigations the images are processed with determination of different number of clusters.
Fig. 4. Main Window of graphical user interface of software tool.
Conclusion. As a result of scientific work the software tool for compression of bitmap images based on clusterization by algorithm of K-means is developed. The elaborated software tool provides the users service to process the optimal compression of selected images in different cases with accounting of proportion between two main parameters – quality of processed image and size of its file. The proposed software tool may be used at different systems: for processing of multimedia information , telecommunication systems, data archiving, security systems, where effective processing of graphics or video sequences is needed.
1. Romanyuk O. N. Computer graphics, Train book, (in Ukrainian), VNTU, FITCE, Vinnitsya, 2021. URL: https://web.posibnyky.vntu.edu.ua/fitki/8romanyuk_komp_grafika/zmg1/ zmg/83.htm
2. Piech Ch. K-Means, 2021. URL: https://stanford.edu/~cpiech/cs221/handouts/ kmeans.html
3. Dabbura I. K-means Clustering: Algorithm, Applications, Evaluation Methods, and Drawbacks, 2020. URL:https:///towardsdatascience.com/k-means-clustering-algorithm-applicati-ons-evaluation-methods-and- drawbacks-aa03e644b48a
4. Jin X., Han J. K-Means Clustering, 2011. In: Sammut C., Webb G.I. (eds) Encyclopedia of Machine Learning. Springer, Boston, MA. https://doi.org/10.1007/978-0-387-30164-8_425
5. Dubuk V.I., Khomenko Yu.V.., Bielan M.S. Development of automated system for control processes of registration of multimedia information (in Ukrainian) // International scientific Inter-net-Conference “Information society: technological, economical and technical aspects of forming (volume 64)”. - Ternopil. - 2021. - P. 18 – 21