Department of Computer Science & Applications

Production of digital images from cameras, mobile phones, camcorders, video films and scanned images has increased exponentially in the past decade. Increasing number of digital devices have influenced human to an extent of taking the same scenes in multiple views. Commercial Organizations, Agencies and Educational Institutions conduct events round the year and the image acquisition activity through the digital devices installed at the event venues is predominant in Universities and Colleges. It is evident that the cost of recording the events is negligible compared to the storage requirement over the years for future reference and inference.  Thus the captured images must be preprocessed to avoid duplication and parsed for low-level processing such as noise removal, contrast enhancement and image sharpening. Few of the processed images can then be chosen based on the weights associated with the event sessions using mid-level processing techniques such as object detection and recognition. Further, the chosen images can be indexed, captioned to be maintained as an archive and stored in the databases to extend quick summary of events for preparing annual documents, editing magazines and departmental newsletters. This paper explores different data mining techniques namely classification and clustering to automatically group the images.

Semantic analysis is a method of representing knowledge. The goal is to reduce the syntactic structures and provide the meaning. English is a language that is geographically accepted all over the world. This is a boon as English is considered to be a language spread worldwide but it could also be a curse because even today people face problem in understanding and accessing this huge corpus of information which is available in English. There is a tremendous amount of information in the internet but all of which are in English. This causes problem for those users who do not understand English. When a word/sentence is translated into another language, it is very important to retain the original meaning even after the translation. This underlying idea has given exposure to Cross language Information Retrieval (CLIR) which is a process of information retrieval in a language different from the language of users query. CLIR is one of the techniques in Natural Language Processing (NLP). In this paper, we discuss about semantic analysis and explore different works that have been done on semantic analysis by different researchers.

As websites on the Internet in the Web 2.0 era have become more interactive, there has been an explosion of new user-generated content. The goal of Summarization Pipeline for Online Repositories of Knowledge (SPORK) is to be able to identify important key topics presented in multi-document texts, such as online comment threads. While most other automatic summarization systems simply focus on finding the top sentences represented in the text, SPORK separates the text into clusters, and identifies different topics and opinions presented in the text. SPORK has shown results of managing to identify 72% of key topics present in any discussion and up to 80%of key topics in a well-structured discussion.

Phonetic analysis is a branch of natural language processing (NLP) which deals with how the sounds are produced when we talk and how words are related to sounds. Applying phonetic analysis to natural language is a challenging task as it involves making computer to understand how sounds are produced and analyze it. Phonetic analysis has its history starting centuries back however over the past few decades the researchers has addressed this problem in a broader perspective concentrating on different languages that are spoken across the world. The biggest provocation for the phonology is to recognize the language, process it for lexical, syntactic, semantic knowledge about the language and the challenges of speech recognition and speech synthesis. In this paper, we present the meaning of phonetics, phonology, its real implementation by investing the features of some of the techniques employed by various researchers.

Implementing phonetics to Natural language is a Herculean task. The first step for the phonetic implementation is to generate a phonetic dictionary. Phonetic dictionary is very important as it plays a vital role in identifying the essential components of the gigantic vocabulary in the speech recognition system of natural language. Language like English needs phonetic transcriptions because the English spelling does not tell us how to pronounce it. Pronunciation is very important for communication as it vitalizes rapid transition to practical concern. There is lot of ongoing research in multilingual speech recognition. Phonetic dictionary is utmost important for any speech recognition system and hence the key goal of the paper is to build a phonetic dictionary for English language. The paper aims at building dictionary for identifying the phonetic transcriptions of International Phonetic Alphabet (IPA) and American English alphabets phonetically. The paper also gives a detailed explanation about the rules of the phonemes.

India has 22 officially recognized languages: Assamese, Bengali, English, Gujarati, Hindi, Kannada, Kashmiri, Konkani, Malayalam, Manipuri, Marathi, Nepali, Oriya, Punjabi, Sanskrit, Tamil, Telugu, and Urdu. Clearly, India owns the language diversity problem. In the age of Internet, the multiplicity of languages makes it even more necessary to have sophisticated Systems for Natural Language Process. In this paper we are developing the phonetic dictionary for natural language processing particularly for Kannada. Phonetics is the scientific study of speech sounds. Acoustic phonetics studies the physical properties of sounds and provides a language to distinguish one sound from another in quality and quantity. Kannada language is one of the major Dravidian languages of India. The language uses forty nine phonemic letters, divided into three groups: Swaragalu (thirteen letters); Yogavaahakagalu (two letters); and Vyanjanagalu (thirty-four letters), similar to the vowels and consonants of English, respectively.

Speaking and writing effectively is probably more important today than never before. Yet while the importance of communication skills increases, many people find it difficult to set about acquiring them. Dictionaries and thesauruses provide a great deal of information and self help material, but their greatest contribution must be restricted to meaning and spelling. With the increasing ambiguity in natural languages, the pressing need is a system that identifies and augments the correct sense to what a word means in a context of usage. Word Sense Disambiguation (WSD) is aimed to achieve the above said goal. There are numerous algorithms in order to disambiguate a word using WSD system such as Lesks’ algorithm, Walkers’ algorithm and Random Walk algorithm. In this paper a complete investigation of lesk algorithm is portrayed. It also includes a detailed analysis of the factors that need to be considered which affect the algorithm. While disambiguating any word in the context of the sentence the inclusive and exclusive of the words is very important and thus we show the grammatical words that creates problem and provide a list of such words that needs to be excluded while evaluating the lesk algorithm.

Uninhabited mistakes while writing happens are unstoppable. There are certain common errors that occur during writing such as missing letters, extra letters, disordered letters, and misspelled letters. These kind of common spelling errors are called phonetics spelling errors. These are of a major concern while dealing with phonetics. Out of various problems that the phoneticians are trying to solve, major portion of it concentrates on varieties of spelling errors. Phonetic structures are greatly emphasized based on the effectiveness, appropriateness and accuracy. In order to keep abreast with the changing and challenging trends of Natural Language Processing (NLP), it is of great importance that one should resolve the problems of spelling errors. To achieve the goal, numerous realistic and practical approaches have to be adopted that make use of spelling correction algorithms such as Edit distance, Habit distance, Soundex and Asoundex. Through the analysis of these algorithms, a new interface is put forward that calculates the Edit distance, thereby showing the overall comparative study of phonetic algorithms with the proposed modified Edit Distance algorithm. The interface computes the Edit distance between two strings in appropriate and intuitive way, contemplating with the comparisons shown in the distance table. The Results show that an average of 0.937 recall and 0.947 precision have been achieved with the F-measure 0.9417. Through these results, it is evident that the recall and F-measures are improved in the proposed Edit-Distance algorithm. The revised version of the edit distance algorithm consistently attains finer quality results in comparison with the traditional edit distance algorithm.

For years now, writing efficiently adapts professional, innovative, unique skills with grammatical knowledge being impeccable. The field of linguistic has its conventional importance especially among anthropologists. In this regard, the Linguistic knowledge has thus proven to be systematic, meticulous, robust and in short, “Scientific Discipline”. The precision in the formulation of new theories hence are envy upon the exact grammatical statements with impressive linguistic descriptions. The English- syntax and linguistic theory has rigorous attempt to construct self-evident detailed level of “Phrase structure” & “Dependency Structure”. These structures are dispensed under Syntactic Representation, a branch in Natural language Processing (NLP). The need for Syntactic anatomy is explained and presents a prototype for “Phrasal, Dependency & Parser Structures”.  In this paper a new algorithm called “Hash Fold Spell Error” (HFSE) using public hash key technique is proposed to arrive at all the possible errors for a given keyword. The obtained results were tested on 2050 words and Recall & Precision was evaluated. It is observed that the algorithm attains high and improved recall and precision in contrast with the existing ones.

Security in network is based on cryptography, the science and art of transforming the messages to make them secure and immune to attack. Cryptography is the most important aspect of communications security, where integrity, non-repudiation, confidentiality, and authentication services plays a vital role .Ciphers are developed to create a secure channel for message communication. The field of cryptography and cryptanalysis is quite demanding and complex as it minimizes the need for time-consuming human interaction with the search process and the network security. However, recurring events such as hackers and intruders attack and the success of criminal attackers illustrate the weaknesses in current security system, information technologies and the need to provide heightened security for these systems. Thus, the application of an efficient and effective tool such as Computational Intelligence (CI) to the field of cryptology comes naturally. CI systems are usually hybrids of paradigms such as Evolutionary Computation systems, Artificial Neural Networks and Fuzzy systems, supplemented with elements of reasoning. Genetic Algorithm is a class of evolutionary computation which is expected to provide optimized and deterministic solution to various common cipher attacks. This paper explores various techniques in cryptography to prove that the natural selection and adaptive mechanism based techniques are as good as the rigorous mathematical techniques used by traditional cryptographic methods.

Security in network is based on cryptography, the science and art of transforming the messages to make them secure and immune to attack. Cryptography is the most important aspect of communications security, where integrity, non-repudiation, confidentiality, and authentication services plays a vital role. Ciphers are developed to create a secure channel for message communication. The field of cryptography and cryptanalysis is quite demanding and complex as it minimizes the need for time-consuming human interaction with the search process and the network security. However, recurring events such as hackers and intruders attack and the success of criminal attackers illustrate the weaknesses in current security system, information technologies and the need to provide heightened security for these systems. Thus, the application of an efficient and effective tool such as Computational Intelligence (CI) to the field of cryptology comes naturally. CI systems are usually hybrids of paradigms such as Evolutionary Computation systems, Artificial Neural Networks and Fuzzy systems, supplemented with elements of reasoning. Genetic Algorithm is a class of evolutionary computation which is expected to provide optimized and deterministic solution to various common cipher attacks. This paper explores various techniques in cryptography to prove that the natural selection and adaptive mechanism based techniques are as good as the rigorous mathematical techniques used by traditional cryptographic methods.

The greatest threats against computer systems around the world come from the cyberspace in the form of cyber-attacks and the zero-day attack is one of such threats faced by computer systems around the world. A vulnerability in system firmware, software, or hardware that is still unknown to the developers or people responsible for it is known as zero-day vulnerability. A zero-day attack is an attack by hackers in which zero-day exploits are applied on a zero-day vulnerability in system firmware, software or hardware before specific security and preventive mechanisms can be identified and set for such vulnerability. This kind of attack is very challenging to defend against because those responsible for the security of such vulnerabilities are unaware of it. Zero-day attacks have been taking a serious upward surge of late and identifying such attacks in real time and quarantining it before it causes more damages to the system is a big problem that computer security personnel are faced with. This research will seek to identify ways in which zero-day attacks can be identified in real time by using a hybrid model that will be proposed. The Signature based defense technique and the Behavior based defense technique are two current security methods that have been identified and this paper will analyze these methods individually and try to find a way in which they can be combined to form a proposed hybrid model.

The greatest threats against computer systems around the world come from the cyberspace in the form of cyber-attacks and the zero-day attack is one of such threats faced by computer systems around the world. A vulnerability in system firmware, software, or hardware that is still unknown to the developers or people responsible for it is known as zero-day vulnerability. A zero-day attack is an attack by hackers in which zero-day exploits are applied on a zero-day vulnerability in system firmware, software or hardware before specific security and preventive mechanisms can be identified and set for such vulnerability. This kind of attack is very challenging to defend against because those responsible for the security of such vulnerabilities are unaware of it. Zero-day attacks have been taking a serious upward surge of late and identifying such attacks in real time and quarantining it before it causes more damages to the system is a big problem that computer security personnel are faced with. This research will seek to identify ways in which zero-day attacks can be identified in real time by using a hybrid model that will be proposed. The Signature based defense technique and the Behavior based defense technique are two current security methods that have been identified and this paper will analyze these methods individually and try to find a way in which they can be combined to form a proposed hybrid model.

The obtained images are frequently flawed because of a variety of environmental issues, particularly at night, such as inside illumination, cloudy weather, etc. The dark image has a compressed dynamic range that can be improved in order to see the finer details. This research effort proposes an improved lighting reflection model-based technique for improving extremely dim images. This improved method relies on reconstruction carried out via morphological processing with Top-hat transformation and Contrast Limited Adaptive Histogram Equalization (CLAHE). The HSV colour scheme is used to perceive the image, and the V component is estimated. The inverse of the intensity component (V) is calculated after normalising the intensity component. The negative image is then subjected to the CLAHE algorithm. The final step is to apply multiscale image enhancement to the obtained image. The brightness component of an image is adaptively adjusted via gamma enhancement. The outcomes of two gamma-enhnced photographs with various gamma values are produced after gamma enhancement. The significant information that can be used for image fusion is extracted from these images using principal component analysis. The weight value is adaptively determined during the PCAbased image fusion employing morphological Top-hat modification to enhance the image quality and highlight the erratic background pixels. The suggested technique emphasises edge and structural preservation while enhancing detail in extremely dim photos. Results from experimental validations demonstrate that the suggested strategy outperforms the current method in terms of both qualitative and quantitative measures.

Digital data including photographs, audio, and video are now readily available
because to the development of the Internet. The ease of access to multimedia raises concerns about ownership identification, authentication of content,
security, and copyright protection. Here, we talk about the idea of digital picture watermarking with an emphasis on the method utilized for embedding and
extracting the watermark from images. This paper also presents a complete
classification of digital watermarking along with its fundamental properties,
such as visual imperceptibility, resilience, capacity, and security. Additionally,
we have covered the most recent uses of digital watermarking in the fields of
healthcare, distance learning, electronic voting, and the military. The robustness is assessed by looking at how image processing assaults affect the content
that is signed and the recoverability of the watermark. The thorough survey
that is offered in this study, in the opinion of the authors, will aid brand-new
scholars in learning more about this area. Additionally, the comparative study
can spark suggestions for how to enhance the methods already outlined.

In this work, we present a novel fuzzy picture segmentation-based watermarking strategy for image copyright protection in the geographical domain. In order to find the best places to hide a watermark, the fuzzy image method uses a features map (edges and regions) that are less perceptible to the human eye. Our approach maximises both imperceptibility and robustness, which are known to be inversely correlated. The PSNR, SSIM, and two similarity metrics based on the fuzzy logic S1, S2' were used to assess the transparency performance of the suggested strategy. The evaluation's findings show how resilient the suggested approach is against popular image processing operators. Because our approach merely makes use of basic processes like fuzzy picture segmentation and image fuzziness, it is also efficient.

Data security has recently become a big concern. Watermarking digital images is viewed as a potential approach for ensuring image data integrity. The existence of various attacks on photos can be detected using watermarking methods. The current study created a watermarking method to secure the integrity and authenticity of medical data and images. Data security in medical has become a vital aspect of hospitals with smart apps to protect privacy of data and security of data which belongs to patients when it comes to photographs. Medical records and diagnostic imaging are typically sent to a specialist physician at another institution for accurate diagnosis. As a result, many academics are intrigued by the idea of transmitting medical data through the internet. Feature Extraction, Image Restoration, Image Enhancement, Embedding, and Extraction are only a few of the digital image processing techniques covered in this work. This paper discusses a brief history of digital images, watermarking for digital images, detecting tampering, picture localization, and restoration.

Digital data including photographs, audio, and video are now readily available because to the development of the Internet. The ease of access to multimedia raises concerns about ownership identification, authentication of content, security, and copyright protection. Here, we talk about the idea of digital picture watermarking with an emphasis on the method utilized for embedding and extracting the watermark from images. This paper also presents a complete classification of digital watermarking along with its fundamental properties, such as visual imperceptibility, resilience, capacity, and security. Additionally, we have covered the most recent uses of digital watermarking in the fields of healthcare, distance learning, electronic voting, and the military. The robustness is assessed by looking at how image processing assaults affect the content that is signed and the recoverability of the watermark. The thorough survey that is offered in this study, in the opinion of the authors, will aid brand-new scholars in learning more about this area. Additionally, the comparative study can spark suggestions for how to enhance the methods already outlined.

The obtained images are frequently flawed because of a variety of environmental issues, particularly at night, such as inside illumination, cloudy weather, etc. The dark image has a compressed dynamic range that can be improved in order to see the finer details. This research effort proposes an improved lighting reflection model-based technique for improving extremely dim images. This improved method relies on reconstruction carried out via morphological processing with Top-hat transformation and Contrast Limited Adaptive Histogram Equalization (CLAHE). The HSV colour scheme is used to perceive the image, and the V component is estimated. The inverse of the intensity component (V) is calculated after normalising the intensity component. The negative image is then subjected to the CLAHE algorithm. The final step is to apply multiscale image enhancement to the obtained image. The brightness component of an image is adaptively adjusted via gamma enhancement. The outcomes of two gamma-enhanced photographs with various gamma values are produced after gamma enhancement. The significant information that can be used for image fusion is extracted from these images using principal component analysis. The weight value is adaptively determined during the PCAbased image fusion employing morphological Top-hat modification to enhance the image quality and highlight the erratic background pixels. The suggested technique emphasises edge and structural preservation while enhancing detail in extremely dim photos. Results from experimental validations demonstrate that the suggested strategy outperforms the current method in terms of both qualitative and quantitative measures. Wang, Wencheng, et al. "Adaptive image enhancement method for correcting low-illumination images". Information Sciences 496 (2019): 25-41.