NCRACS 2017 - National Conference on Recent Advances in Computer Sciences

"NCRACS- Computer Vision 2017 Conference Papers "

Pages   [1] [2] [3] [4] [5] [6]
 

DETECTING IMAGE MANIPULATION USING DEEP LEARNING IN TENSORFLOW[ ]


Multimedia is one of the principal means of communication in these days. When an image or video is obtained as evidence, it can be used as probative onMultimedia is one of the principal means of communication in these days. When an image or video is obtained as evidence, it can be used as probative only if it is authentic. Determining the authenticity of these kinds of multimedia such as images has been an active research area for past few years. Here we propose a system using deep learning algorithm with Tensor Flow as back end, to identify the forgeries on an image. In this paper, we use prediction error filter that mends the altered relationship among the pixels of the image. Experimental results showed that manipulations like median filtering, Gaussian blurring, resizing and cut and paste forgery can be detected with an average accuracy of 96%.

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
QUALITY ASSESSMENTOF SCREEN CONTENT IMAGES[ ]


Research on Screen Content Images (SCIs) has become important as they are used frequently in multi-device communication applications. Perceptual quality is assessed for distorted SCIs subjectively and objectively. A screen image quality assessment database (SIQAD) consisting of 20 source and 980 distorted SCIs is constructed. A subjective quality scores is used for estimating which part (text or picture) contributes more to the overall visual quality. The single stimulus methodology with 11 point numerical scale is used in subjective method. Objective metric is used to measure the visual quality of distorted SCIs which follows subjective analysis. In objective quality analysis a weighting strategy method is used to correlate quality of the two regions (text and pictorial) of SCI and provide quality for the entire image.

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
A HOMOGENEOUS SEMANTIC IMAGE RETRIEVAL USING RELEVANCE BASED WEIGHT ADJUSTMENT TECHNIQUE[ ]


SBIR is a semantic gap between the high-level image and the low-level image. In other words, there is a difference between what image features can distinguish and what people perceives from the image. Semantic Image retrieval is the most complex process in the real time scenario where the similarity finding would be more difficult in case of larger homogeneous image contents. This paper proposes a fuzzy logic based feedback weight adjustment scheme which will increase the score of the class which is more preferred by the users. Through theoreticalanalysis and extensive result, it is shown that our weight adjustment scheme algorithm provides significant performance improvement in terms of high retrieval precision and good user satisfaction level.

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
DETECTION OF BLURRED REGION IN IMAGES FOR FORENSIC APPLICATIONS[ ]


Due to the arrival of new technologies and devices, the crime rate is increasing in developing and developed countries. One such crime is image forgery which can be detected by forensic applications. We propose an idea to identify forgery attack done by blur artifact. In this method, the Region of Interest (RoI) is identified using Histogram Thresholding that involves computation of statistical and color texture analysis. For each RoI, the degree of blur is estimated for distinguishing forged blur artifact from normal blur artifact. The technique is tested by MICC-F220 dataset using MATLAB R2011a. To validate the identified forged blur artifact, we use Fourier and Gabor texture features

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
MULTISCALE PIXEL LEVEL IMAGE FUSION[ ]


Image fusion is a process which combines the data from two or source images from the same scene to generate one single image containing more precise details of the scene than source image. Multistage pixel-level image fusion is a transform coefficient of an image associated with a feature of its value is influenced by the feature’s pixel.Most of the previous image fusion method aim at obtaining as many information from the different modality images. With respect to satellite image fusion the edges and outlines of interested objects is more important than other information. Image with high contrast contain more edge-like features. In our new system we defined the ratio of maximum detail components to the local mean of the corresponding approximate component.

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
BVM SYSTEM USING FINGER PRINT, PHONIC &TELEGENIC APPERCEPTION FOR ENFEEBLE PEOPLE [ ]


The authentication based Biometric System is highly secured and economical to use in the election. Information Technology plays a far-reaching role in recent years. This technology is more annex against offline dictionary attack. The user-id and password mechanism are hired with image processing tactic. The thumb pattern of eligible voters is stored in certain database. During election process, the thumb pattern of voters is taken from the finger print sensor and then the pattern is related with the stored database. For Deaf and groping people the head phones and videos are implemented. The main motive of this System is to overcome the old process of Voting that includes Ballet paper and Punched card with this new Electronic Technology. Finger print recognition system is a new gravitated to enroot security.

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Pages   [1] [2] [3] [4] [5] [6]