Research Cell: An International Journal of Engineering Sciences

Pooja ¹, Renu Dhir ¹

Affiliations
1 Dr. B. R. Ambedker National Institute of Technology, Jalandhar, IN

Abstract

This paper presents a text region extraction approach based on feature analysis. This scheme consists of 5 main phases that are Frame Extraction, Edge Detection and Binarization, Fusion, Normalization and Feature Extraction. For Frame extraction MATLAB simulator is used where video reader is used to extract frames from the video according to the length of video. Edge detection and Binarization is processed on that extracted frames then both the detected edges and binarized image will be fused using PCA based fusion method. Fused image is then normalize and feature extraction is perform by using SIFT and SURF based algorithm. The results will show that extracted text region from the frames of the news videos.

Keywords

Text Region Extraction, Sobel, Otsu Threshold, PCA, SIFT, SURF.

Full Text

   

Dilraj Kaur ¹, Pooja ¹

Affiliations
1 Punjab Technical University, Dept. Computer Science Engineering, CT-Institute of Engineering Management and Technology, Jalandhar, IN

Abstract

Visibility restoration refers to various ways that aim to reduce and remove the degradation that have occurred while the digital image has been obtained. The degradation may be due to various factors like relative object-camera motion, blur due to camera misfocus, relative atmospheric turbulence and others. Underwater image enhancement based algorithms become more useful for many vision applications. It is found that most of the existing researchers have neglected many issues i.e. no technique is accurate for different kind of circumstances. The existing methods have neglected the use of ant colony optimization to reduce the noise and uneven illuminate problem. The main objective of this paper is to evaluate the performance of Ant colony optimization over the available MIX-CLAHE technique.

Keywords

Underwater Haze Removal, ACO, Mix-CLAHE, Dark Channels.

Full Text

   

Gurpreet Kaur ¹, Pooja ¹

Affiliations
1 Punjab Technical University, Dept. of CSE, CT Institute of Engineering, Management and Technology, Jalandhar, IN

Abstract

Generally, color constancy is a method of which calculates the particular impact involving discrete light sources using a digital image. The exacting figure documents with a camera determined by several components such - the actual physical information with the picture, the particular illumination occurrence within the scene, and the features with the digital camera. Our paper has explained the key purpose to modify saturation weighting dependent color constancy applying fuzzy membership based color improvement in addition to edge preserving filtering. As fuzzy membership based saturation weighting may be slow up the impact of the light due to it decreases the sharpness on the image and results in an amount of noise. Therefore a conflict has been arisen and we have needed to remove by using an integrated effort of the edge based color constancy with the histogram stretching, and edge preserving filtering. Experimental results have been shown the efficiency of our approach which carried out certain performance matrices such - MSE, RMSE and PSNR. Our proposed method has given better performance on the basis of performance matrices as compared to the previous methods.

Keywords

Color Constancy, Saturation Weighting, Edge Based Color Constancy, Fuzzy Membership.

Full Text

   

Amandeep Kaur ¹, Pooja ¹, Manju Bala ¹

Affiliations
1 Department of CSE, CT Group of Institutions, Jalandhar, Punjab, IN

Abstract

Use of the decorated text is highly preferable in the area where the attention of viewers is required. The basic optical recognition systems do not give good results for the decorated inputs because the formation of decorated text is complex and different from the other regular fonts. Very less work has been done in the field of decorated text and specially decorated words. In this paper algorithm has been designed which work for the recognition of decorated words. The proposed system uses the neural network for the recognition and consists of the following portion. First is Otsu's algorithm, then applying preprocessing, neural network and finally the matched decision is displayed.

Keywords

Text Recognition, Neural Network, Otsu’s Algorithm.

Full Text

   

Sultana Kadri ¹, Pooja ¹, Manju Bala ¹

Affiliations
1 Department of CSE, CT Group of Institutions, Jalandhar, Punjab, IN

Abstract

In this paper we are proposing a system for the colored saliency detection for securing the systems from indruders. The proposed system is based on saliency detection. The detection Saliency can be a good technique for the detection of Intruders in the security system areas. Saliency detection has become a very prominent subject for research in time. The different types of techniques has been defined for the saliency detection. The experimental results are shown in the below.

Keywords

Saliency Detection, SIFT.

Full Text

   

Sanjeev Dhawan ¹, Kulvinder Singh ¹, Pooja ¹

Affiliations
1 Department of Computer Science, University Institute of Engineering & Technology (U.I.E.T), IN

Abstract

Social network provides number of applications such as Myspace, Facbook, Twitter and many more through which users can connect with their friends and share their images and videos with them. Instagram is application of social network which is used to share images and videos with friends also tag a friend on an image and video. It is difficult to recognize which user is normal user or which user is malicious user. In this paper different techniques to recognize fake profile user has been surveyed and provide a mechanism to detect fake profiles in social network. This paper proposed a mechanism to detect normal posts using Random Forest classifier. The proposed mechanism has been analyzed using weka.

Keywords

Social Networks, Fake Profile, Cloning ,Instagram and Facebook.

Full Text