A color filter array is placed on the image sensor of a digital camera to acquire color images. Each pixel uses only one color, since the image sensor can measure only one color per pixel. Therefore, empty pixels are filled using an interpolation process called demosaicing. The original and the interpolated pixels have different statistical characteristics. If the image is modified by manipulation or forgery, the color filter array pattern is altered. This pattern change can be a clue for image forgery detection. However, most forgery detection algorithms have the disadvantage of assuming the color filter array pattern. We present an identification method of the color filter array pattern. Initially, the local mean is eliminated to remove the background effect. Subsequently, the color difference block is constructed to emphasize the difference between the original pixel and the interpolated pixel. The variance measure of the color difference image is proposed as a means of estimating the color filter array configuration. The experimental results show that the proposed method is effective in identifying the color filter array pattern. Compared with conventional methods, our method provides superior performance.
We propose an efficient Markov feature extraction method for color image splicing detection. The maximum value among the various directional difference values in the discrete cosine transform domain of three color channels is used to choose the Markov features. We show that the discriminability for slicing detection is increased through the maximization process from the point of view of the Kullback–Leibler divergence. In addition, we present a threshold expansion and Markov state decomposition algorithm. Threshold expansion reduces the information loss caused by the coefficient thresholding that is used to restrict the number of Markov features. To compensate the increased number of features due to the threshold expansion, we propose an even–odd Markov state decomposition algorithm. A fixed number of features, regardless of the difference directions, color channels and test datasets, are used in the proposed algorithm. We introduce three kinds of Markov feature vectors. The number of Markov features for splicing detection used in this paper is relatively small compared to the conventional methods, and our method does not require additional feature reduction algorithms. Through experimental simulations, we demonstrate that the proposed method achieves high performance in splicing detection.
We present a blind identification of image manipulation types such as blurring, scaling, sharpening, and histogram equalization. Motivated by the fact that image manipulations can change the frequency characteristics of an image, we introduce three types of feature vectors composed of statistical moments. The proposed statistical moments are generated from separated wavelet histograms, the characteristic functions of the wavelet variance, and the characteristic functions of the spatial image. Our method can solve the n-class classification problem. Through experimental simulations, we demonstrate that our proposed method can achieve high performance in manipulation type detection. The average rate of the correctly identified manipulation types is as high as 99.22%, using 10,800 test images and six manipulation types including the authentic image.
The light information obtained through a color filter array is sampled as R (red), G (green), and B (blue) based on the intensity of the radiation sensor CCD charge-coupled device (CCD) or complementary metal oxide semiconductor (CMOS). Because one pixel in the CCD or CMOS has a wave value for only one color, the information on the other two colors has to be estimated from the samples of the neighboring pixels. This estimation process is called demosaicking, and it plays an important role in the image pipeline. We propose a demosaicking method that uses the correlation among the wavelet coefficients in the high-frequency region of the color channels (R, G, and B). In our method, the high-frequency wavelet coefficients are obtained by linear estimation. The low-frequency coefficients are replaced by the results obtained from bilinear interpolation. We simulate our method and compare it with existing demosaicking schemes. The experimental results show that the proposed method can perform good demosaicking.
We develop a spatially adaptive Bayesian image denoising method using a mixture of a Gaussian distribution and a point mass function at zero. In estimating hyperparameters, we present a simple and noniterative method. We use a hypothesis-testing technique in order to estimate the mixing parameter, the Bernoulli random variable. Based on the estimated mixing parameter, the variance for a clean signal is obtained by using the maximum generalized marginal likelihood (MGML) estimator. We simulate our denoising method using both orthogonal wavelet and dual-tree complex wavelet transforms and compare our algorithm to well-known denoising schemes. Experimental results show that the proposed method can generate good denoising results.
We present a new deinterlacing algorithm based on modularization by the local frequency characteristics of images. The input patterns of an image are divided into two regions—the edge region and the smooth region. Each region is assigned to one neural network. The local frequency characteristics of patterns are similar within each region, resulting in more accurate training for each network. The regional neural networks are composed of two modules. One is for the low-frequency components of the input pattern, and the other is for the high-frequency components. Both modules are combined in the output. Therefore, each module compensates for the drawbacks of the other. In simulation, the proposed algorithm showed better performances in both still images and video sequences than other algorithms.
When images compressed with block-based compression techniques are transmitted over a noisy channel, unexpected block losses occur. Conventional methods that do not consider edge directions can cause blocked blurring artifacts. In this paper, we present a post-processing-based block recovery scheme using Haar wavelet features. The adaptive selection of neighboring blocks is performed based on the energy of wavelet subbands (EWS) and difference between DC values (DDC). The lost blocks are recovered by linear interpolation in the spatial domain using selected blocks. The method using only EWS performs well for horizontal and vertical edges, but not as well for diagonal edges. Conversely, only using DDC performs well for diagonal edges with the exception of line- or roof-type edge profiles. Therefore, we combine EWS and DDC for better results. The proposed directional recovery method is effective for the strong edge
because exploit the varying neighboring blocks adaptively according to the edges and the directional information in the image. The proposed method outperforms the previous methods that used only fixed blocks.
In general, wavelet coefficients are composed of a few large coefficients and a lot of small ones. Therefore, each wavelet coefficient is efficiently modeled as a random variable of a Bernoulli-Gaussian mixture distribution with unknown parameters. The Bernoulli-Gaussian mixture is composed of the multiplication of the Bernoulli random variable and the Gaussian mixture random variable. In this paper, we propose a denoising algorithm using the Bernoulli-Gaussian mixture model based on sparse characteristics of the wavelet coefficient. The denoising is performed with Bayesian estimation. We present an effective denoising method through simplified parameter estimation for the Bernoulli random variable using a local expected square error. Simulation results showed that our method outperformed the states of the art denoising methods.
Deinterlacing is the conversion process from the interlaced scan to progressive one. While many previous algorithms that are based on weighted-sum cause blurring in edge region, deinterlacing using neural network can reduce the blurring through recovering of high frequency component by learning process, and is found robust to noise. In proposed algorithm, input image is divided into edge and smooth region, and then, to each region, one neural network is assigned. Through this process, each neural network learns only patterns that are similar, therefore it makes learning more effective and estimation more accurate. But even within each region, there are various patterns such as long edge and texture in edge region. To solve this problem, modular neural network is proposed. In proposed modular neural network, two modules are combined in output node. One is for low frequency feature of local area of input image, and the other is for high frequency feature. With this structure, each modular neural network can learn different patterns with compensating for drawback of counterpart. Therefore it can adapt to various patterns within each region effectively. In simulation, the proposed algorithm shows better performance compared with conventional deinterlacing methods and single neural network method.
A wavelet coefficient is generally classified into two categories: significant (large) and insignificant (small). Therefore, each wavelet coefficient is efficiently modelled as a random variable of a Gaussian mixture distribution with unknown parameters. In this paper, we propose an image denoising method by using mixture modelling of wavelet coefficients. The coefficient is classified as either noisy or clean by using proper threshold [2]. Based on this classification, binary mask value that takes an important role to suppress noise is produced. The probability of being clean signal is estimated by a set of mask values. Then we apply this probability to design Wiener filter to reduce noise and also develop the method of selecting windows of different sizes around the coefficient. Despite the simplicity of our method, experimental results show that our method outperforms other critically sampled wavelet denoising schemes.
Wavelet transform is used efficiently for high compression ratio image coding. It is useful for a high resolution and color document image processing system. However, a large amount of memory is required in wavelet decompression for a large image. Conventional wavelet transform does not allow reconstructing a sub-region of the image. Therefore, it is hard to use wavelet transform in a color document image processing system. In this paper, a block wavelet transform method is proposed. An image is divided into blocks and the samples of the adjacent block are used to transform one block for removing the edge artifact. The required number of samples of the adjacent block are derived. By using the proposed method, the memory requirement for wavelet coding/decoding is reduced to that of the block size. Any targeted region of an image can be compressed or reconstructed. The proposed method can be used for a color document image processing system.
In this paper, we propose a new image compression technique using wavelet transform and human visually estimated noise sensitivities. These consist of frequency, background brightness, and edge height sensitivities. The background brightness sensitivity for each quantizing point is modeled by a quadratic function. The edge sensitivity for each quantizing point is modeled by a non-linear function. The minimum value becomes background brightness sensitivity and edge height sensitivity is multiplied by the frequency sensitivity for determining the quantization step size. Quantization step sizes are calculated by using coefficients of lowest frequency band which are coded losslessly. Therefore, in the proposed method, information to specify quantization step size for higher frequency band, is not needed. The coefficients of high frequency bands are arithmetically coded in horizontal and vertical directions depending on the edge direction. Compared with previous human visual systems based image compression methods, the proposed method shows improved image quality for the same compression ratio with less computational cost.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.