Semantic Driven Automated Image Processing Using ... - Science Direct

Available online at www.sciencedirect.com

ScienceDirect Procedia Computer Science 58 (2015) 453 – 460

Second International Symposium on Computer Vision and the Internet (VisionNet’15)

Semantic Driven Automated Image Processing using the Concept of Colorimetry Merugu Suresha*, Kamal Jainb a*

Indian Institute of Technology Roorkee, Uttatakhand,247667, India. [email protected]. Indian Institute of Technology Roorkee, Uttatakhand,247667, India. [email protected].

b

Abstract

Color is the visual perceptual property of humans. Color obtains from the band spectrum light-weight {of sunshine} (dissemination of wavelength versus light energy) interacting within the eye with the spectral band sensitivities of the sunshine receptors. Color classification and actual descriptions of color are also connected with objects, materials, lightweight sources, etc., supported their physical properties like light-weight absorption, reflection, or emission spectra. Colorimetry (quantitative chemical analysis) is that the science and psychophysical aspects of color that describes colorise numbers or provides a physical color match employing a kind of measuring instruments. Ontologies can be used and designed to

reproduce dispense knowledge and control the semantic heterogeneity among color domains. Knowledge is collected by ability of individuals and represented in description logics. Thus ontology is a key implementation strategy of semantic research, which provides knowledge footprint of the corresponding domain. This paper presents ontology for using various color categories. The resulting ontology can be used by human or machine to extract the relevant prominent features from different color categories to find out the inter-connection among physical aspects of color model and therefore the physiological aspects of human vision. This paper is basically an approach to automate the digital image processing method using the concept of colorimetry and semantics. This paper focused on the development of semantic information which represents the color domains of the image resolution pixel value. The aim of this ontology is to modify machines to create and perceive visual color description schemes and descriptors that are expressed in web ontology language (OWL) format. © 2015 2015 The Elsevier B.V. This is an open access article under the CC BY-NC-ND license © The Authors. Authors.Published Publishedbyby Elsevier B.V. (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of organizing committee of the Second International Symposium on Computer Vision and the Peer-review under responsibility of organizing committee of the Second International Symposium on Computer Vision and the Internet Internet(VisionNet’15). (VisionNet’15) Keywords: Colorimetry, ontology, OWL, color images, color models and sub pixel analysis

1. Introduction In recent years, a huge number of Images have been uploaded on the web e.g., high resolution images, large scale image archives, distributed color image repositories, etc; Contents in images are highly complicated and

1877-0509 © 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of organizing committee of the Second International Symposium on Computer Vision and the Internet (VisionNet’15) doi:10.1016/j.procs.2015.08.062

454

Merugu Suresh and Kamal Jain / Procedia Computer Science 58 (2015) 453 – 460

semantically unknown. Currently, image automated feature extraction focus on the classification of low-level or minute concepts, detection, and identification of relevant static features in an image. These tasks are typically accomplished by exploiting attributes like texture, shape, structure. Images, as snapshots require Ontologies include semantics associated with this paper and it discusses the semantic retrieval of image pixel domain color models. The ontology based semantic information is an augmentation of the current worldwide web where data is well understood that acts as an interface between machine and users for performing information, image and video retrieval. Semantic Web Technologies like Ontology provides a new way image of annotation for describing image features Huang (2004). Ontological analysis of the pixel domains of images has highlighted the different perspectives that exist about the nature of an image pixel. To build color domain and graphical knowledge representation of procedure between heterogeneous knowledge resources here every one should require repeatable ontology construction process in the semantic integration of ontology technology. The changing face of Image Enhancement processing has profoundly digitized the older pixel classification color system. This is at the hamlet of multiple domains from biology to geology; and from textile to printing industries. In this new Era several companies hold legacy color models and emulate for their mark power and their potential to envision how a color will be used on a massive difference of materials (paper, products, plastics, textiles and various substances used in interior and exterior design) (from crpit.com). Presently world has already achieved digital reform where all electronic devices (all computing devices in Virtually), database is constituted in a digital format using different spatial arrangements of binary information (bits) that can encode numbers within finite constraints and with some minimum precision. By variance, the color images we perceive on this earth habitat are comparable to another. These are formed by compound influences between day light and actual objects, materials resulting in perpetual variations in light (band) wavelength and intensity (reflectance value) (Suresh and Jain, 2011 & 13). For example, captured an image and represent it on any device or a computer display e.g. with a scanned data or metric camera (the display device complement of an eye). Since this should not have unlimited repository (bits), it follows that user must convert this comparable signal (image descend) into a more finite digital in format. This computation process called as image (signal) sampling. Sampling (down sampling or up sampling) theory is an important part of Computational processing and Computer Vision and Graphics. Here one should simply observe that a digital color image cannot convert into a coded form absolute exceptional amount of detail, nor absolutely vast (‘dynamic’) color domain. Preferably, one must understand on issues of accuracy, by selecting a suitable algorithm to descend the signal and reserve (i.e. constitute) our perpetual color image in a digital constitution. The Computer Vision and Graphics aerosol to the main issue of color image in digital form is to split the color image (picture) up into a well organized grid that can call a ‘raster’. Each grid cell is a ‘picture element’, a term often engaged to pixel. The pixel is an atomic unit of the image; it is colored uniformly its single color representing a separate sample of light e.g. from a metric image.

Fig.1. Raster are used to represent digital color images and this shows grey scale frame buffer in raster and the intensities of the pixels (0=black, 255=white).

In most digital image processing, raster data takes the form of a regular straight line grid consists of thousands of pixel elements (shown in the fig. 1). The number of pixel elements in a color image is referred to as the image’s spatial resolution. Present desktop device displays (LCDs and LEDs) are proficient of visualizing images with

Merugu Suresh and Kamal Jain / Procedia Computer Science 58 (2015) 453 – 460

455

spatial capability of sensors around 1024 × 768 pixels (i.e. a million pixel elements or one mega-pixel). Generally, the greater the spatial resolution, greater the level of spatial resolution of color image representation. For all Different types of output devices, such as monitors, printers, scanners etc, and the most important thing is to show the right information to the user. Pixel is the basic element both on-screen and materials printed using devices. And, as a result pixel color processing is the key technique to make the output correct, precise, and suitable to use on different outcomes. However XML database imparts the acceptable, constitutional constraints required for Color classification domains, this does not give the semantic interoperability required to make image pixel as visual descriptors accessible by other domains. The machine learning knowledge representation provided in the form of ontology can be used to develop a tool which introduces knowledge based reasoning in color classification domain. Even this supports flexibility for carrying out various applications of multimedia tasks. 2. Related work (Kanimozhi, 2013) He gave the overview of how to make SPARQL query and incorporating Ontology, which acts as an interface between the system and humans for providing data information, image and video retrieval. Web Ontology Language (OWL), widely used to construct domain ontology and for graphical representation of Ontology, OntoViz and OntoGraph were used which improves the efficiency of data information search. (John, 1995) developed a new approach for automatically single/multiple color extractions and indexing to support color queries of images as well as video databases, which specifies both color content and spatial characteristics of image regions and visual features like size and shape. (Athanasiadis, Simou et al. 2009) they have proposed and implemented a methodology for ontology based semantic indexing of images using the digital image processing techniques like image classification, image segmentation and fuzzy knowledge based reasoning to extract the semantic description of raw content information which is used for indexing and retrieval of images. 3. Structure of Ontology An initial working and development stage of Color Ontology, Unified Modeling Language (UML) was used to model the class properties and relationships which comprised different models. Ontology accession and prototype modifications are unfavourable for designers of models of systems wherever the challenge is that the linguistics integration of heterogeneous resources into a domain-rich and customer-focused network-centric application (from crpit.com, 2015; Anuj, 2015; Simou, 2005; and Berners, 2001). Color is that the brains reaction to a selected visual input. Though the inadequacy of heritage model hinders the development of Color ontology, it also means that the generated ontology will be even more valuable, providing a definition of the semantic information of color models and the semantic relationships between them. Here building the color ontology should also highlight any incompatibilities, ambiguities or duplication, which exists across the maximum number of Color descriptors. The color model example what has shown in the form of algorithm and demonstrates the process of comparatively better resources relating totally different systems of references(from crpit.com). Pixel (PIX [picture] Element) the smallest addressable unit displayed on Output screens. The higher the value of pixel closure (the more rows and columns of pixels), the more selective information can be displayed. Each pixel may be composed of sub pixels. For example a pixel on a color display may be composed of Cyan, Magenta, Yellow, Black sub-pixels. A pixel in a video signal may be composed of LAB parts or any other color models. 3.1. Class and Data Properties This section describes all ontology classes and their data properties. Color is the parent class for all classes depicted in fig. 2. It is implemented as an abstract class. Color class has following properties: ColorLayoutDescriptor: property of type int represents Layout of image;

456

Merugu Suresh and Kamal Jain / Procedia Computer Science 58 (2015) 453 – 460

ColorQuantizationDescriptor: property of type string represents quality of image; ColorStructureDescriptor: property of type int represents Structure of image; CMYK: property of type int represents color models of an image; DominantColorDescriptor: property of type string represents Dominant color of image; HSB: property of type int represents color model value of an image pixel; HSL: property of type int represents color model value of an image pixel; LAB: property of type int represents color model value of an image pixel; RGB: property of type int represents color model value of an image pixel; ScalableColorDescriptor: property of type int represents scale and range of image; XYZ: property of type int represents color model value of an image pixel;

Fig.2. Various Color Classes and color models

3.2. Important Individual Sub-classes of Color class This section describes all important individual Sub-class of Color class depicted in fig. 3. 1) CMYK: This class has ranges of color space values, and can be called as CMJN. The colorimetric color space ranges from 0 to 100% in most applications or values (0 to 255). It is standard color model for scanning and printing. The fourth, black component is included to improve the density range. 2) HSB: This class has ranges of color gamut be made up of the most useful components like hue, saturation and brightness. Which have obtained from the RGB color model. 3) HSL: This class has ranges of wealth color gamut be made up of the components like hue, saturation and lightness. 4) LAB: This class has ranges of color gamut which has the components like luma, green/red and blue/yellow. 5) RGB: This class color model depends on trichromatic color gamut often found in system that uses a CRT display images. It is commonly used almost in all computer display devices as well as television, video etc. 6) XYZ: is root of all colorimetry and with these values chromaticity coordinates (x, y) can be derived.

Merugu Suresh and Kamal Jain / Procedia Computer Science 58 (2015) 453 – 460

457

Fig.3. Color Models

4. Ontology enabled colorimetry A simple ontology for color and their relationships to other components is depicted in fig. 5. Our Color ontology contains classes (almost exclusively named color domains) and individuals (instances of colors such as CMYK, HSV, LAB, RGB, intervals and HTML codes and Coordinates), developed using Protégé Ontology Tool Editor. Maximum correlations between feature classes are currently shown through the “hasCoordinate” correlations described by the Region Connection Calculus. HTML Codes can also be used to define as instances of the Colors. Color hasProperty color Structure descriptor, dominant Color Descriptor, Scalable Color Descriptor, Color Quantization, Color Models, Color Layout Descriptor, and Color Space Descriptor (Wen, 2012). Each Color Models IsDividedInto CMYK, HSB, HSL, LAB, RGB, XYZ. 5. Class relation Class relations describe relations between individual ontology classes. Most important classes and relations are: HasProperty: Describes the relation between color and Descriptors. HasCoordinate: Describes the relation between color and color space value. IsDividedInto: Describes the relation between color and colormodels. 5.1. Algorithm Procedure Render Image () For each pixel on the Image Move or place Cursor C from the view position of the Image. Lists the Coordinates of Color Models of a particular pixel. Call the procedure OWL Color Modes () with Action Listener of Mouse with arguments X, Y and Z. Plot the pixel color value Coordinates returned by Coordinate Trace (). Next pixel Position End Procedure Fig.4. Showing the flow of logic algorithm

458

Merugu Suresh and Kamal Jain / Procedia Computer Science 58 (2015) 453 – 460

Here color picker works by spontaneous represent the color category of the pixel right below your mouse curser pointer as you progress mouse curser over image pixel. During this model mouse curser-drag selecting model comes in. One will click on the actual points of image pixel to a pixel you wish to represent, and once your mouse curser pointer is over the required place, release the mouse button to pick the color value and can see its color variation.

Fig.5. Color Ontology OWL describing the relations between different color models

5.2. Querying color domains: Once defined the color ontology for a particular class, one have to be compelled to outline a technique for querying the information data from the database. SPARQL is that the candidate recommendation of W3C for querying Ontology and designed specifically to support semantic web applications (Vincent, 2008) and information. Fig. 6 shows the SPARQL code to find out the color gamut for all color models.

Fig.6. Query for RGB color Models

Merugu Suresh and Kamal Jain / Procedia Computer Science 58 (2015) 453 – 460

a

459

b

d

c

Fig.7. From (a) to (d) Query for various color Models and color variations using colorimetry from a color image

a

b

c

Fig.8. From (a) to (c) showing the transformed image from the original image and with color variations in (x, y) chromaticity diagram for the color palate

6. Conclusion From many centuries philosophers, researchers and scientists have enough explored the domain of digital image processing and the nature of color. More recently, philosophers have enriched the debate about Automatic methods for searching image collections that make wide use pixel level image processing methods that are robust to large changes in viewpoint, and can be computed trivially. However, older methods like histograms fail to incorporate semantic information, and therefore tend to give vague results. This paper introduces a new domain of colorimetry to extract the actual descriptions of color related with objects, materials, light sources, etc. When results is presented over the platform of semantic web and implemented as ontology strength the path of automated decision making tools in digital image processing for image annotation and searching. The resulting ontology can be used by human or machine to extract the relevant prominent features from different color categories to find out the inter connection and ontology enabled structure allow to understand the interdependencies among physical aspects of color model and therefore the physiological aspects of human vision. SPARQL was used to extract the knowledge with respect to the user perception to show the relationship between class and subclass, object and data properties and annotated with relevant image color model. Integration of ontology and SPARQL for semantic web improves efficiency of image color models. References: 1.

Anuj Tiwari, Janga Reddy, Kamal Jain, “semantic evolution in publication system”, Cognitive Computing and .Information Processing CCIP - 2015, (3 - 4 March 2015) Noida, IEEE Xplore, http://www.ieee.org.

460

Merugu Suresh and Kamal Jain / Procedia Computer Science 58 (2015) 453 – 460 2. 3. 4. 5.

6. 7. 8. 9. 10. 11. 12.

13. 14. 15.

Anuj Tiwari, P. Srujana, K. Rajesh, “semantically enriched knowledge extraction with data mining”, International Journal of Computer Applications Technology and Research (IJCATR), 2015, Volume 04, Issue 01, pp-07-10, Jan 2015, ISSN: 2319–8656. Anuj Tiwari, Dr. Kamal Jain, “ontology based local spatial data infrastructure for smart city development”, 15th Esri India User Conference id: UCP0043, 09-11 December 2014, Delhi, India. Berners-Lee, T., Hendler, J., & Lassila, O. “The semantic web”, Scientific American, 2001, 284(5), 28-37. Huang, W., & Webster, D., September. “Enabling Context-Aware Agents to Understand Semantic Resources on The WWW and The Semantic Web”. In Proceedings of the 2004 IEEE/WIC/ACM International Conference on Web Intelligence (pp. 138-144). IEEE Computer Society. John R. Smith and Shih-Fu Chang, “Single Color Extraction and Image Query” International Conference on Image Processing (ICIP95), Washington, DC, Oct. 1995. Kanimozhi. T, Dr. A. Christy, “Incorporating Ontology and SPARQL for Semantic Image Annotation”, proceedings of IEEE Conference on Information and Communication Technologies (ICT 2013). Lefort, L. and Taylor, K,. “Large scale colour ontology generation with XO”. In Proc. Australasian Ontology Workshop (AOW 2005), Sydney, Australia. CRPIT, 58. Meyer, T. and Orgun, M. A., Eds. ACS. 47-52. Merugu Suresh, Kamal Jain, “Development of Sub-Pixel Analysis Method Based on Colorimetry”, 2-Day International Conference on Advanced Computing Methodologies, in Elsevier, ICACM-2011, organized by GRIET, Hyderabad, during December 9-10, 2011. Merugu Suresh, Kamal Jain, “Sub Pixel Analysis on Hypothetical Image by using Colorimetry”, International Journal of Recent Technology and Engineering (IJRTE), ISSN: 2277-3878, Volume 2, Issue-4, and September 2013. Simou N, Tzouvaras. V, Avrithis. Y, Stamou. G, Kollias. S, “A Visual Descriptor Ontology for Multimedia Reasoning”, 2005, Workshop on Image Analysis for Multimedia Interactive Services (WIAMIS 2005); 01/2005 Thanos Athanasiadis, Nikos Simou, Georgios Th. Papadopoulos, Rachid Benmokhtar, Krishna Chandramouli, Vassilis Tzouvaras,Vasileios Mezaris, Marios Phiniketos, Yannis S. Avrithis, Yiannis Kompatsiaris, Benoit Huet, Ebroul Izquierdo, “Integrating Image Segmentation and Classification for Fuzzy Knowledge-Based Multimedia Indexing”, Advances in Multimedia Modeling, 15th International Multimedia Modeling Conference, MMM 2009, Sophia-Antipolis, France, January 7-9, 2009. Proceedings; 01/2009. Vincent Huang. "Semantic Sensor Information Description and Processing", 2008 Second International Conference on Sensor Technologies and Applications (sensorcomm 2008), 08/2008. Wen Yong-ge. "Research on Image Retrieval Based on Scalable Color Descriptor of MPEG-7", Lecture Notes in Electrical Engineering, 2012. www.crpit.com (The Conference in Research and Practice in Information Technology (CRPIT) series (a parallel series to the Journal of Research and Practice in Information Technology)), visited on 20th April 2015.