Suah Kim

Reversible Watermarking Based on Compensation

This paper proposes a high performance reversible watermarking (RW) scheme based on a novel compensation strategy. RW embeds data into a host image by modifying its pixel values slightly. It is found that certain modified pixels can be compensated to their original values during the proposed embedding procedure. The compensation effect in the RW scheme can improve the marked image quality significantly. By incorporating the pixel selection method, a higher quality image is obtained, which is verified by extensive experiments.

Linear collaborative discriminant regression classification for face recognition

A highly discriminative sub-space learning method is proposed.A novel collaborative between-class reconstruction error is maximized.The small class-specific between-class reconstruction error is emphasized.Linear regression classification performs well in the learned sub-space. This paper proposes a novel face recognition method that improves Huangs linear discriminant regression classification (LDRC) algorithm. The original work finds a discriminant subspace by maximizing the between-class reconstruction error and minimizing the within-class reconstruction error simultaneously, where the reconstruction error is obtained using Linear Regression Classification (LRC). However, the maximization of the overall between-class reconstruction error is easily dominated by some large class-specific between-class reconstruction errors, which makes the following LRC erroneous. This paper adopts a better between-class reconstruction error measurement which is obtained using the collaborative representation instead of class-specific representation and can be regarded as the lower bound of all the class-specific between-class reconstruction errors. Therefore, the maximization of the collaborative between-class reconstruction error maximizes each class-specific between-class reconstruction and emphasizes the small class-specific between-class reconstruction errors, which is beneficial for the following LRC. Extensive experiments are conducted and the effectiveness of the proposed method is verified.

Automatic contrast enhancement using reversible data hiding

Automatic image enhancement is increasingly becoming a popular tool for the smartphone environment. The tool automatically enhances the image right after it has been stored to enhance users experience. But, because enhancements are subjective and dependent on the image, the original image is backed up to provide a recovery option. This requirement inevitably increases the storage requirement. In order to reduce it, a novel automatic contrast enhancement based on reversible data hiding (ACERDH) is proposed. The proposed method mimics the equalization effect observed in the basic contrast enhancement technique called global histogram equalization, while providing reversibility. The experiment visually show improved contrast. Additional experiment was done to compare the embedding capacity with an another reversible data hiding based contrast enhancement technique [4]. The proposed method is fit for automation, while providing data hiding capability and removing the additional storage requirement.

Breaking character and natural image based CAPTCHA using feature classification

ABSTRACT CAPTCHA(Completely Automated Public Turing test to tell Compute rs and Humans Apart) is a test used in computing to distinguish whether or not the user is computer or human. Many web sites mostly use the character-based CAPTCHA consisting of digits and characters. Recently, with the development of OCR technology, simple character-based CAPTCHA are broken quite easily. As an alternative, many web sites add noise to make it harder for recognition. In this paper, we analyzed the most rec ent CAPTCHA, which incorporates the addition of the natural images to obfuscate the characters. We proposed an efficient method using support vector machine to separate the characters from th e background image and use convolutional neural network to recognize each characters. As a result, 368 out of 1000 CAPTCHA s were correctly identified, it was demonstrated that the current CAPTCHA is not safe.Keywords: CAPTCHA, Breaking CAPTCHA, SVM, CNN, HSV color space * 접수일(2015년 5월 4일), 수정일(1차: 2015년 7월 14일,2차: 2015년 8월 3일), 게재확정일(2015년 8월 19일)*This work was supported by the National Research Foundation of Korea Grant funded by the Korean government(MEST) (NRF-2015R1A2 A2A01004587) and supported by Business for

Reversible Data Hiding Based on Combined Predictor and Prediction Error Expansion

This paper presents a novel reversible data hiding method which uses a combined predictor. The proposed combined predictor combines five base predictors according to their global and local predicting performance. The weights to combine the base predictors are calculated with a pixel by pixel manner that they adjust to the local image patch characteristics. The proposed predictor is shown to have high prediction precision which is beneficial for the following prediction error expansion (PEE). Observing that our predictor performs well even for images with complex textures, a novel pixel selection criterion that bases on the prediction errors is proposed, which can accurately select the pixels that have small prediction errors to use. Extensive experiments are conducted to verify the superior performance of the proposed method.

Reversible watermarking using edge based difference modification

Reversible watermarking can embed data into the cover image and extract data from stego image, where the original cover image can be recovered perfectly after the extraction of data. Difference expansion (DE) and prediction error expansion (PEE) are two popular reversible watermarking methods. DE has the advantage of small distortion while PEE has the advantage of large embedding capacity and smaller prediction error compared with pixel difference. In this paper, we proposed a novel method that combines the advantages of DE and PEE, where the difference calculated between two pixels is combined with the edge information near this pixel pair. The proposed difference calculation can produce smaller pixel difference compared with the original simple pixel difference calculation. Overlapping embedding is then used to increase the embedding capacity. Our proposed method gives excellent results which is shown by several experiments.

Weighted sparse representation using a learned distance metric for face recognition

This paper presents a novel weighted sparse representation classification for face recognition with a learned distance metric (WSRC-LDM) which learns a Mahalanobis distance to calculate the weight and code the testing face. The Mahalanobis distance is learned by using the information-theoretic metric learning (ITML) which helps to define a better weight used in WSRC. In the meantime, the learned distance metric takes advantage of the classification rule of SRC which helps the proposed method classify more accurately. Extensive experiments verify the effectiveness of the proposed method.

Analyzing the effect of the distortion compensation in reversible watermarking

Reversible watermarking is used to hide information in images for medical and military uses. Reversible watermarking in images using distortion compensation proposed by Vasily et al [5] embeds each pixel twice such that distortion caused by the first embedding is reduced or removed by the distortion introduced by the second embedding. In their paper, because it is not applied in its most basic form, it is not clear whether improving it can achieve better results than the existing state of the art techniques. In this paper we first provide a novel basic distortion compensation technique that uses same prediction method as Tian’s [2] difference expansion method (DE), in order to measure the effect of the distortion compensation more accurately. In the second part, we will analyze what kind of improvements can be made in distortion compensation.