Showing papers by "V. N. Manjunath Aradhya published in 2018"

A Novel Arbitrary-Oriented Multilingual Text Detection in Images/Video

Abstract: Text in images and videos plays a vital role to understand the events. The textual information is a prominent source and semantic information of a particular content of the respective image or video. Text detection is a primary stage for text recognition and text understanding. Still, text detection process is a challenging and interesting research work in the field of computer vision due to illumination, alignments, complex background and variation size, color, fonts of the text. The multilingual text consists of different geometrical structures of languages. In this paper, a simple and yet effective approach is presented to detect the text from arbitrary oriented multilingual images/video. The proposed method is based on Laplacian of Gaussian information and full connected component analysis. The proposed method is evaluated on four datasets such as Hua’s dataset, arbitrarily oriented dataset, Multi-script Robust Reading Competition (MRRC) dataset and MSRA dataset with performance measures precision, recall and f-measure. The results show that the proposed method is promising and encouraging.

Color Features and KNN in Classification of Raw Arecanut images

Abstract: Arecanut is one of the important cash crops of Southern India. Classification of raw arecanut is one of the major tasks in grading, which is a vital part of crop management. In this work we proposed a model which classifies the raw arecanut. We used color histogram and color moments as features with K-NN classifier. Experiment is conducted on a dataset of 800 images of four classes using two color features and four distance measures with K-NN. A classification accuracy of 98.13% is achieved for 20% training with K value of 3 and Euclidean distance measure for color histogram features.

Classification of sentiments in short-text: an approach using mSMTP measure

Abstract: Sentiment analysis or opinion mining is an automated process to recognize opinion, moods, emotions, attitude of individuals or communities through natural language processing, text analysis, and computational linguistics. In recent years, many studies concentrated on numerous blogs, tweets, forums and consumer review websites to identify sentiment of the communities. The information retrieved from social networking site will be in short informal text because of limited characters in blogging site or consumer review websites. Sentiment analysis in short-text is a challenging task, due to limitation of characters, user tends to shorten his/her conversation, which leads to misspellings, slang terms and shortened forms of words. Moreover, short-texts consists of more number of presence and absence of term/feature compared to regular text. In this work, our major goal is to classify sentiments into positive, negative or neutral polarity using new similarity measure. The proposed method embeds modified Similarity Measure for Text Processing (mSMTP) with K-Nearest Neighbor (KNN) classifier. The effectiveness of the proposed method is evaluated by comparing with Euclidean Distance, Cosine Similarity, Jaccard Coefficient and Correlation Coefficient. The proposed method is also compared with other classifiers like Support Vector Machine and Random Forest using benchmark dataset. The classification results are evaluated based on Accuracy, Precision, Recall and F-measure.

LoG and Structural Based Arbitrary Oriented Multilingual Text Detection in Images/Video

Abstract: Text in an image or a video affords more precise meaning and text is a prominent source with a clear explanation of the content than any other high-level or low-level features The text detection process is a still challenging research work in the field of computer vision However, complex background and orientation of the text leads to extremely stimulating text detection tasks Multilingual text consists of different geometrical shapes than a single language In this article, a simple and yet effective approach is presented to detect the text from an arbitrary oriented multilingual image and video The proposed method employs the Laplacian of Gaussian to identify the potential text information The double line structure analysis is applied to extract the true text candidates The proposed method is evaluated on five datasets: Hua's, arbitrarily oriented, multi-script robust reading competition (MRRC), MSRA and video datasets with performance measures precision, recall and f-measure The proposed method is also tested on real-time video, and the result is promising and encouraging

Segmentation of Brain Tumor Tissues in HGG and LGG MR Images Using 3D U-net Convolutional Neural Network

Abstract: The quantitative assessment of tumor extent is necessary for surgical planning, as well as monitoring of tumor growth or shrinkage, and radiotherapy planning. For brain tumors, magnetic resonance imaging MRI is used as a standard for diagnosis and prognosis. Manually segmenting brain tumors from 3D MRI volumes is tedious and depends on inter and intra observer variability. In the clinical facilities, a reliable fully automatic brain tumor segmentation method is necessary for the accurate delineation of tumor sub regions. This article presents a 3D U-net Convolutional Neural Network for segmentation of a brain tumor. The proposed method achieves a mean dice score of 0.83, a specificity of 0.80 and a sensitivity of 0.81 for segmenting the whole tumor, and for the tumor core region a mean dice score of 0.76, a specificity of 0.79 and a sensitivity of 0.73. For the enhancing region, the mean dice score is 0.68, a specificity of 0.73 and a sensitivity of 0.77. From the experimental analysis, the proposed U-net model achieved considerably good results compared to the other segmentation models.

eLL: Enhanced Linked List—An Approach for Handwritten Text Segmentation

Abstract: In document image analysis, the interesting and challenging lie in textline segmentation of handwritten documents. This work emphasizes on developing an enhanced handwritten textline segmentation technique based on the concept of linked list. The proposed system consists of three phases namely preprocessing, enhanced linked list (eLL), and morphology processing. The experiment is evaluated on a document containing handwritten Kannada and English script, and the results are promising.