ICTACT Journal on Image and Video Processing

R. Krishnamoorthi ¹, T. Karthikeyan ²

Affiliations
1 Department of Information Technology, Bharathidasan Institute of Technology, Trichy, Tamil Nadu, IN
2 Department of Computer Science and Engineering, PSG College of Arts and Science, Coimbatore, Tamil Nadu, IN

Abstract

In order to replace the artifacts in the textured background, a new texture preserving image coder using the set of orthogonal polynomials is proposed in this paper. The proposed scheme is based on the model that represents textures using points spread operator relating to a linear system. In the proposed texture based image coding scheme, the encoder first identifies textured regions, which are then analyzed to produce the model features. Then these features are later transmitted to decoder which decodes to form a synthetic texture and results into synthetic stage. The proposed modeling delivers to attain high compression ratio by maintaining constantly excellent visual quality. 92.31% with a PSNR value of 31.93dB when the quality factor is 5 for D96 image is achieved by the proposed scheme. By keeping up the quality factor as a constant constrained, we obtain 91.11% of compression ratio with a PSNR value of 33.26dB for different set of image that is, D38 image.

Keywords

Texture Preserving Image Coder, Points Spread Operator, Synthetic Texture, Texture Modeling and Compression Ratio.

Full Text

     

T. Karthikeyan ¹, R. Krishnamoorthy ²

Affiliations
1 Department of Computer Science, PSG College of Arts and Science, IN
2 Department of Computer Science and Engineering, Anna University, Tiruchirappalli, IN

Abstract

In this study autoregressive model based on Bayesian approach is proposed for texture classification. Based on auto correlation coefficients, micro textures are identified and represented locally and then globally. The identified micro texture is represented as a local description, called texnum. The global descripter, texspectnum, is obtained by simply observing the numbers of occurrences of the texnums that cover the entire image. The proposed representation scheme has been employed in both supervised and unsupervised classifications of textured images. The supervised classification is based on simple tests of hypotheses and the unsupervised classification is based on the modified K-means algorithm with minimum distance classifiers. The proposed method is demonstrated for classification of different types natural textured images. The average correct classification is better than the existing methods.

Keywords

Texnum, Texspectnum, Microtexture, K-Means Algorithm, Supervised and Unsupervised Classification.

Full Text

     

M. Ramkumar ¹, Amirtha Preeya ², R. Manikandan ³, T. Karthikeyan ⁴

Affiliations
1 Department of Computer Science and Engineering, HKBK College of Engineering, IN
2 Department of Computer Science and Engineering, Presidency University, IN
3 Department of Computer Science, The Quaide Milleth College for Men, IN
4 Department of Electronics and Telecommunications Engineering, University of Technology and Applied Sciences, OM

Abstract

In this paper, we develop a liveness detection of iris present in the study to reduce various spoofing attacks using gray-level co-occurrence matrix (GLCM) and Deep Learning (DL). The input images of iris are been given to this technique for the extraction of texture and colour features with fine details. The details are fused finally and given to a DL classifier for the classification of liveness detection. The simulation is conducted to test the efficacy of the model and the results of simulation shows that the proposed method achieves higher level of accuracy than existing methods.

Keywords

Iris Detection, Pattern Identification, Liveness Detection, Biometric, Deep Learning

Full Text