http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
- 中文 을 입력하시려면 zhongwen을 입력하시고 space를누르시면됩니다.
- 北京 을 입력하시려면 beijing을 입력하시고 space를 누르시면 됩니다.
RISS 인기검색어
- 검색키워드
- 저자 : Sonia Carole Kouayep (검색결과 4 건)
- 무료
- 기관 내 무료
- 유료
-
An Enhanced Copy-Move Forgery Detection based on Radon Binary Pattern
Sonia Carole Kouayep,Kyung-Hyune Rhee 한국정보통신학회 2014 2016 INTERNATIONAL CONFERENCE Vol.6 No.1
Recently blind image forensics which aims to assess the image authenticity has been raised a great awareness in multimedia security. One of the most popular methods among this category is copy-move forgery detection. This paper proposes an enhanced algorithm for detecting such forgeries in digital images. To deploy this approach, firstly, we divide the image into some overlapping blocks and then capture the intensity of the pixels in each block by utilizing the Radon Transform. Secondly to provide the computational efficiency of the Radon domain, we encode the micro-information of image to the Local Binary Pattern as feature vectors. Finally, in matching step, we use the Chi-square histogram-based distance to determine similar or duplicated blocks. The experimental results show that the proposed joint descriptors for extracting features vectors are very effective in accurate detection of copied regions even when these regions have undergone lossy compression. The detection accuracy and performance of our method in comparison to similar works, promising to be efficient even in detection rate.
-
Applications of Multiple Instance Learning - A Review
Samman Fatima,Kouayep Sonia Carole,Sikandar Ali,Hee-Cheol Kim 한국정보통신학회 2023 한국정보통신학회 종합학술대회 논문집 Vol.27 No.1
Multiple Instance learning is gaining more and more popularity over the recent years because of the flexibility in its application, as it fits different and special scenarios. It is basically a variation of supervised learning. Contrary to supervised learning where every instance is properly labelled with a discrete value, MIL supports weakly labelled data with incomplete information. Each instance in training data is not assigned a discrete label rather a single label is assigned to a set of instances known as bag. MIL is applied in many real-life scenarios to solve business problems where labelling data is expensive like medical imaging, computer vision, time series and document classification etc. There are many algorithms for MIL. In this paper we will briefly describe the scenarios where MIL is applicable.
-
Detection of Copy-Move Image Forgery Based on SIFT-LBP
Kueh Lee Hui,Sonia Carole Kouayep 한국정보기술학회 2015 한국정보기술학회논문지 Vol.13 No.7
Here, we proposes a new view of an enhanced copy-move forgery detection using Local Binary Pattern (LBP) based on Scale Invariant Features Transform (SIFT). This framework computes rotating invariant subuniform local binary patterns from the image keypoints. The SIFT algorithm is first applied to the converted grayscale image from the original image to extract scale invariant keypoints of the image. Then, the subuniform LBP is computed centered at the keypoints to extract the feature vector of each keypoint. Finally, Chi-square histogram-based distance is computed for matching purpose to determine similarity. Besides that, Random Sample Consensus (RSC) algorithm is adopted in to this framework to remove mismatches. Our experiment results show that the proposed method can produce accurate detection results and exhibit high robustness to scale and rotation forged regions.
-
Blind Image Detection Based on Fusion Feature
Kueh Lee Hui,Sonia Carole Kouayep 한국정보기술학회 2016 한국정보기술학회논문지 Vol.14 No.4
Here, we proposed the fused feature based on the theoretical analysis and experimental verification of Local Binary Pattern (LBP) operator and Level Co-occurrence Matrix (GLCM) features to provide robust and complementary features than a single descriptor set for improving the detection performance of the blind image. This framework was focused on whether the image is blind or not by applying the fusion of two texture descriptor algorithms. The LBP algorithm is firstly employed to the image to create LBP image (LBP features). Then, the GLCM algorithm is applied to the LBP image to extract local co-occurrence features. The fused features both gain spatial distribution from the LBP image and the co-occurrence features. Finally, Support Vector Machine (SVM) is computed for forgery detecting on the image. The experiment results show that the proposed framework can achieve the accuracy of 98.5% on color image dataset of CASIA database.
연관 검색어 추천
이 검색어로 많이 본 자료
활용도 높은 자료
