-In this paper, the idea of creating a system to analyze and improve quality of 2D images is presented. Proposed model operates on self-organizing Kohonen network. For this purpose, the method of image processing and preparation of vectors representing the components of the image are described. Tests on various images were made and presented.
I. INTRODUCTION 2D image processing is not only a very important part of today’s science but ubiquitous technology. Mobile phones, police speed cameras, or analysis of images in different factories are just a few basic applications of processing of 2D graphics. This is the main motivator for creating new and improving existing methods of detection and analysis of shapes, or improve quality of graphics.
For the image analysis, it is important to prepare the image,
in a certain way. For this purpose, a number of filters are used
to minimize the amount of information contained in the image
leaving only the essential information or delete a plurality
of noise and distortion. An example of a filter is a filter for
removing noise with using the theory of fuzzy sets, and other
methods of artificial intelligence [
An important aspect of the image processing is their
compression. File compression increases possibility of easy
transfer and processing of files. Compression algorithms should
not only reduce the weight of the file, but keep the best image
quality. One example of the newer compression algorithms
that use rbfnn and discrete wavelet decomposition is shown
in [
Copyright c 2016 held by the author.
One of the most popular applications of image processing is
medicine - the detection of various diseases in the early stages
can save lives. In [
An interesting topic in the field of image processing are
neural networks, which are often used in the classification
of different objects or even the entire image. In [
Quick and effective methods have numerous applications in factories, wineries and orchards eg.: in [26] was shown the algorithm for detecting defects of fruit based on pictures using radial basis probabilistic neural networks.
In this work, I would like to introduce an idea of a system to improve image quality. For this purpose, an innovative way to extract data about the image quality from image file is discussed. In addition, implementing self-organizing Kohonen network to indicate what needs to be improved in order to get the best quality picture is presented.
II. HSL
HSL next to RGB and CMYK is one of the most famous models of color space. It was first presented in [27] as a model associated with the perception of color by the human eye. Each color is perceived as a light coming from a certain point (lightning), what is more, each color is derived from white light. The name comes from the proposed model of the three characteristics of color Hue, Saturation and Lightness.
Color model HSL is understood as a cone in which the color wheel is the base of the cone (see Fig. 1). Each color can be represented as a three-element vector of the following form [h; s; l] ; where all values describe one component of the cone.
The hue h is understood as the angle on the color wheel which takes a value between h0 ; 360 i. The color wheel begins with the red color and subsequently at 120 moves to a different color (120 is green and 240 is blue). Formally, the hue is a property that the human eye can classify as one of three primary colors (red, green, blue). Determination of the hue occurs according to h = 8 >>60 > > > <
60 > > > >>:60
G
B
(mod6) if B R
R G + 2 + 4 if if = R = G = B where = max(R; G; B) and = min(R; G; B). In the case where = 0, it is considered that the value is indeterminate.
The second value describing the HSL model is a saturation that is described as the radius of the base that takes values of h0; 1i. The formula describing this attribute is s = 1 j + where = then s = 0.
The third and the last variable of the model is lightness l. It is interpreted as the height of the cone. Lightness as well as saturation takes values in the range h0; 1i. It is defined as the average value of the largest and smallest components of color what can be represented as the following formula + l = :
2
III. KOHONEN’S SELF-ORGANIZING MAP
The first models of artificial neural networks have already appeared in the 40s of the twentieth century [28]. More than 30 years later, a Finnish scientist Teuvo Kohonen has developed a model of neural networks that learning does not require supervision [29], [30]. Applied learning is called competitive learning or learning with the competition. After entering patterns on the network, winning neuron is determined only this one neuron and its neighborhood have updated weight. In the case of this type of network, an important element is the choice of distance measure. With this measure, the network creates image of topological space of the input signals.
Euclidean metric is the most common metric. The mathematical formula between two points x1 and x2 is defined as d(x1; x2) = jjx1
vu n x2jj = tuX(x1i i=0 x2i)2; where n is the number of point coordinates.
Learning operates by selection of the winning neuron which the weights are similar to the input vector. It can be represented by d(x; wn) =
min (d(x; wi)): i=1;2;;n (1)
Using the selected metric, the size of the neighborhood is chosen. The radius of the neighborhood is reduced with successive epochs. In the next step, the weight of the selected neurons are updated by the following equation wi(t) = wi(t 1) + f (i; x)(x wi(t 1)); where is a learning parameter, t is the number of epoch and f (i; x) is a function of the neighborhood defined as a Gaussian function as follows g(i; x) = e (d(i; w))2
The proposed system consists of two parts – preparation of the vector representing the image and Kohonen classifier.
The system accepts a 2D image, which is divided into four parts. Then, the six points of (x; y) are selected at random. Points must be within a smaller area of the image. For each image, the selected points are found. Then, the neighborhood of 12 points is determined for each point. The arithmetic average of each value (hue, saturation, lightness) is calculated for all areas defined by the neighborhood. As a result, four vectors are created. All of the vectors are combined in a single thirteen-element vector representing the quality of the input image. Created vector takes the following form [h1; h2; h3; h4; s1; s2; s3; s4; l1; l2; l3; l4; c]; (9) where c is the value of 1 when the image is correct, and 0 in 8 >>>>hi = > > > > > > <
si = > > > > > > >>>li = > : where i means the number of the image, j is the number of neighborhoods and k is the total number of points in the neighborhood.
The resulting vector can be added to the database or be assessed by Kohonen classifier. The system classifies the 2D image in terms of its quality. In the case where any of the components of the HSL model differ from the norm, this component should be improved by increasing/decreasing according to mathematical formulas in Sec. II. In the case of learning, vectors stored in database are used. A model of such system is illustrated in Fig. 3.
In order to test the proposed system, 100 pictures were taken – 80 pictures with a digital camera with a resolution of 15 Mpx and 20 images capture with the camera in a mobile phone with a resolution of 8 Mpx. Among the samples of photos taken with a digital camera, 50 of them were made in good quality. All images taken with the camera in a mobile phone were made in the best quality.
For each photo, the vector was created according to the notation in (9). Then, the vectors were added to the database with corresponding markings c. Kohonen classifier has been made of two layers: an input (13 neurons) and output (9 9 grid of neurons). Learning on the network is performed using all samples from the database to achieve 10000 epochs. In order to know the percentage of correctness of the network we check the results for all samples in the database. The obtained result was 71% correctly classified images.
The results of the network considered and applied the appropriate corrections by increasing or decreasing eg.: saturation. Sample images before and after improvement are shown in Fig. 4 and 5.
Algorithm 1 Kohonen Network Algorithm 1: Start 2: Initiate a learning parameter , Gaussian function parameter , the number of epochs 3: Set weights at random 12: 15: 13: 14: 10: 11: 4: while t < epochs do 5: for each input vector x do 6: for each output neuron k do 7: Calculate the distance according to (5) 8: Find the winning neuron using (6) 9: Find the neighborhood of the winning neuron for each neuron in the neighborhood do
Update the weight using (7) end for
Reduce the radius of the neighborhood end for end for 16: end while 17: Stop
The presented model not only allows to determine whether a particular 2D image is correct in terms of quality, but it shows what needs to be improved. Implementation of algorithms to improve the quality of basic HSL attributes is performed by specific equations shown in Sec. II for each pixel of the image.
Result correctness of the analyzed samples were obtained at the level of 71%, which is a good result due to the small number of photos taken in the learning process. The system is unable to cope with the correct classification of the images in the most complex color (eg.: partially obscured), which is its disadvantage.
In the future research work is planned to consider a more complex system in terms of execution time, learning time and more parameters than the HSL model.
Author acknowledge contribution to this project of Operational Programme: Knowledge, Education, Development financed by the European Social Fund under grant application POWR.03.03.00-00-P001/15. [21] Z. Marszałek and M. Woz´niak, “On possible organizing nosql database systems,” Int. J. Information Science and Intelligent System, vol. 2, pp. 51–59, 2013. [22] R. K. Nowicki, B. A. Nowak, J. T. Starczewski, and K. Cpalka, “The learning of neuro-fuzzy approximator with fuzzy rough sets in case of missing features,” in Neural Networks (IJCNN), 2014 International Joint Conference on. IEEE, 2014, pp. 3759–3766. [23] R. K. Nowicki, M. Korytkowski, B. A. Nowak, and R. Scherer, “Design methodology for rough neuro-fuzzy classification with missing data,” in Computational Intelligence, 2015 IEEE Symposium Series on. IEEE, 2015, pp. 1650–1657. [24] W. K. Mleczko, T. Kapus´ci n´ski, and R. K. Nowicki, “Rough deep belief network-application to incomplete handwritten digits pattern classification,” in Information and Software Technologies. Springer, 2015, pp. 400–411. [25] B. A. Nowak, R. K. Nowicki, M. Woz´niak, and C. Napoli, “Multi-class nearest neighbour classifier for incomplete data handling,” in Artificial Intelligence and Soft Computing. Springer, 2015, pp. 469–480. [26] G. Capizzi, G. Lo Sciuto, C. Napoli, E. Tramontana, and M. Wozniak, “Automatic classification of fruit defects based on co-occurrence matrix and neural networks,” in Computer Science and Information Systems (FedCSIS), 2015 Federated Conference on. IEEE, 2015, pp. 861–867. [27] G. H. Joblove and D. Greenberg, “Color spaces for computer graphics,” in ACM siggraph computer graphics, vol. 12, no. 3. ACM, 1978, pp. 20–25. [28] W. S. McCulloch and W. Pitts, “A logical calculus of the ideas immanent in nervous activity,” The bulletin of mathematical biophysics, vol. 5, no. 4, pp. 115–133, 1943. [29] C. Von der Malsburg, “Self-organization of orientation sensitive cells in the striate cortex,” Kybernetik, vol. 14, no. 2, pp. 85–100, 1973. [30] T. Kohonen, “Self-organized formation of topologically correct feature maps,” Biological cybernetics, vol. 43, no. 1, pp. 59–69, 1982.