Low Frequency DCT Research Articles

Robust Image Hashing via CP Decomposition and DCT for Copy Detection

Copy detection is a key task of image copyright protection. This article proposes a robust image hashing algorithm by CP decomposition and discrete cosine transform (DCT) for copy detection. The first contribution is the third-order tensor construction with low-frequency coefficients in the DCT domain. Since the low-frequency DCT coefficients contain most of the image energy, they can reflect the basic visual content of the image and are less disturbed by noise. Hence, the third-order tensor construction with the low-frequency DCT coefficients can ensure robustness of our algorithm. Another contribution is the application of the CP decomposition to the third-order tensor for learning a short binary hash. As the factor matrices learned from the CP decomposition can preserve the topology of the original tensor, the binary hash derived from the factor matrices can reach good discrimination. Lots of experiments and comparisons are done to validate effectiveness and advantage of our algorithm. The results demonstrate that our algorithm has superior classification and copy detection performances than several baseline algorithms. In addition, our algorithm is also better than some baseline algorithms with regard to hash length and computational time.

DCT based multi-head attention-BiGRU model for EEG source location

Electroencephalogram source imaging (ESI) pertains to localize brain sources. Due to the one-to-many relationship between electroencephalogram (EEG) signals and brain sources, ESI becomes a complex inverse problem with poor conditioning, leaving a lot of research space. This article proposes an enhanced multi-head attention (MA) and discrete cosine transform (DCT) based Bidirectional Gated Recurrent Unit (BiGRU) (MA-DCT-BiGRU for short) for addressing the EEG inverse problem. Initially, EEG signal characteristics are captured employing multi-head attention with the inclusion of average hidden states. Then, the DCT is used to project brain source signals into the low, medium, and high frequency subspaces composed of spatial frequency basis vectors. The spatial low-frequency component serves as a filter for extended source reconstruction. Subsequently, BiGRU is employed to learn the mapping from the output of the attention layer to the low frequency DCT coefficients of the brain-derived signals. The simulation results undeniably establish the superiority of the MA-DCT-BiGRU configuration in comparison to other state-of-the-art (SOA) methods for source recovery, irrespective of the source modes and signal-to-noise ratio conditions. Experimental results, utilizing both synthetic and actual epilepsy data, clearly demonstrate the effectiveness of the framework presented in this article for epileptogenic area localization.

Theoretical foundations for constructing effective codewords for the code-controlled information embedding steganographic method

Steganography is an important component of modern information security systems. At the same time, in the conditions of modern cyberspace, it is relevant to develop high-performance steganographic methods that would have a high level of resistance to possible attacks by compression, noise, and blur. One of such methods is the steganographic method with code-controlled information embedding, based on the idea of preliminary coding of the information being embedded using binary codewords, for which the transformants of the Walsh-Hadamard transform have the specified properties. A specific localization of disturbances in the Walsh-Hadamard transform domain of the container takes place because of the information embedding. In this paper, a theoretical basis has been formed for further improvement of the codewords used in the code-controlled information embedding steganographic method. It is shown that despite the fact that these codewords have an ideal effect only on a given transformant of the Walsh-Hadamard transform, they affect several transformants at once in the domain of the discrete cosine transform (DCT). The concept of the selectivity coefficient is introduced to estimate the level of selectivity of the impact on a given DCT transformant. It has been established that with an increase in the size of the blocks used, a tendency is observed to a decrease in the selectivity coefficient due to the presence of the “close neighbor” effect. This trend is conditioned by the involvement of the DCT transformants with similar frequencies that have similar resistance to possible attacks on the embedded message. In this case, the ratio of the sum of absolute values of low-frequency DCT transformants to the sum of absolute values of all other DCT transformants increases with the size of the codeword. In this paper it has been proven and practically confirmed that an increase in the size of a codeword leads to an increase in the resistance of the code-controlled information embedding steganographic method. Possible ways of further practical improvement of codewords used in the code-controlled information embedding steganographic method are theoretically substantiated.

Perceptual image hashing using transform domain noise resistant local binary pattern

A new Discrete Cosine Transform (DCT) domain Perceptual Image Hashing (PIH) scheme is proposed in this paper. PIH schemes are designed to extract a set of features from an image to form a compact representation that can be used for image integrity verification. A PIH scheme takes an image as the input, extracts its invariant features and constructs a fixed length output, which is called a hash value. The hash value generated by a PIH scheme is then used for image integrity verification. The basic requirement for any PIH scheme is its robustness to non-malicious distortions and discriminative ability to detect minute level of tampering. The feature extraction phase plays a major role in guaranteeing robustness and tamper detection ability of a PIH scheme. The proposed scheme fuses together the DCT and Noise Resistant Local Binary Pattern (NRLBP) to compute image hash. In this scheme, an input image is divided into non-overlapping blocks. Then, DCT of each non-overlapping block is computed to form a DCT based transformed image block. Subsequently, NRLBP is applied to calculate NRLBP histogram. Histograms of all the blocks are concatenated together to get a hash vector for a single image. It is observed that low frequency DCT coefficients inherently have quite high robustness against non-malicious distortions, hence the NRLBP features extracted from the low frequency DCT coefficients provide high robustness. Computational results exhibit that the proposed hashing scheme outperforms some of the existing hashing schemes as well as can detect localized tamper detection as small as 3% of the original image size and at the same time resilient against non-malicious distortions.

A novel non-linear modifier for adaptive illumination normalization for robust face recognition

In this paper, a novel approach is presented for adaptive illumination normalization for face recognition under varying illuminations due to change in angle of light projection. Illumination normalization is performed over some of the low frequency discrete Cosine transform (DCT) coefficients which are computed adaptively based upon the significance of alteration of these coefficients. These are, then, modified using a non-linear modifier. The significance of the proposed approach is that the approach is adaptive in normalizing the illumination from the face images as the number of low frequency DCT coefficients that need to be modified using non-linear modifier are selected on the basis of variations of illumination present in the image. This variation in the illumination is also used in developing the non-linear modifier. The proposed approach is important in such a way that the level of illumination variations decides the computations needed i.e. lesser in comparison to the other existing state-of-art approaches as large number of frequency coefficients are not altered. The proposed approach is tested over various face databases: YALE B, Extended YALE B, CMU PIE, AR and YALE face database. The experimental results clearly reveal that the performance of the proposed approach is significantly better than the existing approaches of illumination normalization for face recognition. With the proposed approach, 100% accuracy is attained on all the subsets of YALE B, CMU PIE and on Subset 3 of Extended YALE B database. Promising results are achieved over remaining face databases as well.

An adaptive illumination normalization using non-linear regression for robust person identification under varying illuminations

The precision of person recognition using facial images in complex illumination conditions deteriorates severely. Various illumination normalization and compensation techniques alleviate the classification performance by linear and static modeling of illumination variations thus result in inappropriate illumination normalization. Here, the authors present novel and adaptive illumination normalization method with the help of kernel extreme learning machine (KELM). Few low frequency discrete-Cosine-transform (DCT) coefficients are used as input to KELM for learning the non-linear regression model for illumination variations. The trained KELM is used to find the illumination variations present in the input face image which is represented by the regression value. Depending upon the regression value corresponding to the illumination variations in the test image, the number of low frequency DCT coefficients is selected adaptively. The selected low frequency DCT coefficients are processed by a fuzzy filter. The processed face images are classified by KELM with polynomial kernel. The evaluation of the proposed illumination normalization method is conducted on Extended Yale B face database. Zero, 0.19 and 1.11 percent error rate on subset 3, 4 and 5, respectively, of this database proves the efficacy of the proposed method.

A Novel Algorithmic Approach for Efficient Realization of 2-D-DCT Architecture for HEVC

Low power realization of high efficiency video coding (HEVC) standard, which has found widespread application in mobile devices, tablets, and high-definition televisions, calls for an efficient approximation of discrete cosine transform (DCT) technique. In this paper, a novel algorithm is proposed to determine the minimum number of low-frequency DCT coefficients required for transform and quantization block in HEVC. Moreover, it introduces hardware efficient 1-D architectures for 4, 8, 16, and 32-point DCT, that make use of the proposed algorithm and conform to the HEVC standard. When these architectures are extended to the HEVC test model (HM), it leads to a minor reduction of PSNR (<;0.15 dB) and a slight increase in bitrate (<;1.5%). The encoding time of HM encoder is reduced up to 16.02% and produces the rate-distortion performance comparable to the reference algorithm. A new memory efficient addressing scheme for transpose memory is presented for 2D-DCT. The proposed 2D-DCT architecture has a low power dissipation and energy consumption of 8.62 mW and 1.77 pJ respectively and can process at least sixty 8K ($7630\times 4830$ ) frames/s. Such architecture with low power and energy characteristics can be integrated into a real-time HEVC encoder for portable consumer electronic products.

Hardware Efficient Architecture for 2D DCT and IDCT Using Taylor-Series Expansion of Trigonometric Functions

This paper presents a hardware architecture for 8 x 8 2D Discrete Cosine Transform (DCT) and Inverse DCT (IDCT) using Taylor-series expansion of trigonometric functions. The processing of DCT/IDCT is modified to eliminate the need for: 1) partitioning the input image into fixed-size blocks and 2) use of transpose buffer for storing intermediate results, achieving low clock cycle time compared to existing techniques. The values obtained from Taylor-series are approximated by fixed-point numbers that reduce the hardware complexity at the cost of acceptable loss in quality of the output image. Based on the fixed-point approach, 18-bit, 24-bit, and 32-bit DCT and IDCT architectures are proposed. This DCT architecture considers the implementation of all 64 coefficients and hence, it is named as DCT64. An important observation about 2D DCT is that most of the compaction energy is concentrated on the low-frequency DCT coefficients which are sufficient to reconstruct the original image. Taking this point into consideration, three DCT architectures, namely, DCT15, DCT10, and DCT6 with 15, 10, and 6 coefficients are designed which are compared with the DCT64. The percentage improvement in the area and performance utilization is obtained for DCT15, DCT10, and DCT6 against DCT64. A gradual increase in the percentage is observed from DCT15 through DCT6 with a marginal decline in the quality of the output image at each step which can be seen from the measured PSNR and SSIM values. Regarding IDCT, the proposed architecture achieved a low percentage of area utilization and similar performance to that of the DCT. The field-programmable gate array (FPGA) implementation of the proposed DCT architecture operates at a higher frequency, achieving lower values of dynamic power, slice reg, slice LUTs metrics that outperforms the processing capabilities of existing Algebraic Integer (AI) and fixed-point implementations.

Image up-sampling using deep cascaded neural networks in dual domains for images down-sampled in DCT domain

Recent researches show that the high-frequency discrete cosine transform (DCT) coefficients can be estimated from low-frequency DCT coefficients by exploiting the spatial correlations. Hence, images coded by DCT such as JPEG/MJPEG/H.264, etc., can be down-sampled in DCT domain, where the high-frequency information can be accurately restored through image up-sampling. In this letter, we propose a novel deep neural network using the cascaded fully connected layers and convolution layers in dual domains (DCT and spatial domains), in order to restore high-frequency DCT coefficients from observed low-frequency DCT coefficients by exploiting the DCT inter-block and spatial correlations. In the proposed network, many recent techniques are adopted, including residual network in dual domains, batch normalization, denseNet, etc. Experimental results show that the proposed cascaded networks in dual domains significantly outperforms the state-of-the-art DCT up-sampling methods in terms of PSNR (0.63–2.57 dB gain), SSIM values, and subjective evaluations on standard image datasets Set5 and Set14.

Dual image watermarking by exploiting the properties of selected DCT coefficients with JND modeling

The distinctive properties of partly sign-altered (PSA) coefficients in discrete cosine transform (DCT) domain are explored to achieve effective dual blind image watermarking. A host image is divided into non-overlapped blocks, each of which is then converted to a DCT representation separately. For each block, low-frequency DCT coefficients are selected for dual binary watermarking. The first binary bit is applied to the mean value of the PSA coefficients. Our formulation takes into account of human psychovisual characteristics. A stair-like quantization function is designed to not only regulate embedding strength but accommodate extra embedding for blocks with high contrast masking. In case second watermarking is requested, we tactically adjust the standard deviation of the PSA coefficients while keeping the PSA mean intact. The embedding strengths with respect to the involved parameters in the first and second watermarking are analytically examined. Experiment results indicate that the proposed scheme is capable of achieving excellent robustness at a peak signal-to-noise ratio (PSNR) around 38.7 dB. In particular, the exploitation of just noticeable distortion helps to enhance the robustness of the first watermark without causing noticeable quality degradation. The secondary watermark is embedded at the cost of approximately 1 dB PSNR. Yet the resultant performance is comparable with other compared methods in terms of bit error rates.

Exploring robust and blind watermarking approach of colour images in DWT-DCT-SVD domain for copyright protection

This paper presents a new robust and invisible blind watermarking approach of colour images for copyright protection in hybrid DWT-DCT-SVD domain. In the proposed method, firstly the luminance component (Y) of the cover image is decomposed up to one level of discrete wavelet transform (DWT) coefficients and the low frequency band (LL) is transformed by discrete cosine transform (DCT). Then several selected low and intermediate frequency DCT coefficients of each block are extracted to generate a feature matrix and singular value decomposition (SVD) transform is applied to the feature matrix. Finally the watermark information is embedded by modifying the singular values of the feature matrix. Experimental results demonstrate that the proposed approach outperforms some popular existing watermarking methods in robustness against median filtering, Gaussian filtering, salt and pepper noise, average filtering, Gaussian noise, histogram equalisation, and so on, especially in case of lossy JPEG compression in addition to good imperceptibility.

Exploring robust and blind watermarking approach of colour images in DWT-DCT-SVD domain for copyright protection

This paper presents a new robust and invisible blind watermarking approach of colour images for copyright protection in hybrid DWT-DCT-SVD domain. In the proposed method, firstly the luminance component (Y) of the cover image is decomposed up to one level of discrete wavelet transform (DWT) coefficients and the low frequency band (LL) is transformed by discrete cosine transform (DCT). Then several selected low and intermediate frequency DCT coefficients of each block are extracted to generate a feature matrix and singular value decomposition (SVD) transform is applied to the feature matrix. Finally the watermark information is embedded by modifying the singular values of the feature matrix. Experimental results demonstrate that the proposed approach outperforms some popular existing watermarking methods in robustness against median filtering, Gaussian filtering, salt and pepper noise, average filtering, Gaussian noise, histogram equalisation, and so on, especially in case of lossy JPEG compression in addition to good imperceptibility.

An improved video key-frame extraction algorithm leads to video watermarking

The epidemic growth of internet services has created a broad attention on digital content authentication and copyright protection. Heading the issue we are going to propose a DCT based novel approach on invisible video watermarking along with an improved video key-frame extraction algorithm. In this article first the key-frames (video bookmarks) of the video sequence are being extracted. In order to identify the video bookmarks, a novel method is proposed based on boundary luminosity analysis that is proven efficient regarding fast camera movement and moving objects in the video frames. A pair of secret keys will be applied during the watermark embedding phase. Our framework embeds watermarks in only those video frames which have been identified as key-frames of that video stream because, the key-frames are mostly kept as it is in different compression algorithms. The low frequency DCT coefficients of key-frames are chosen for watermarking, that again offers robustness to the compression algorithms. Instead of embedding watermark bits directly, a new method is proposed- scaled average. The watermark is perceptually invisible in the video key-frames according to human visual system. At last uncompressed video is re-reconstructed with all frames of the video in appropriate order of appearance. The watermark extraction is blind i.e., except the keys nothing is needed for watermark extraction.

Robust Image Watermarking Based On Quantization Index Modulation in the DCT Domain

The main drawback of the traditional quantization index modulation (QIM) watermarking is its extreme sensitivity to scaling attack. Due to scaling have a little effect on Discrete Cosine Transform (DCT) coefficients, this paper proposed a novel robust image watermarking algorithm based on QIM in the DCT domain. In the presented scheme, the medium and low frequency DCT coefficients of host signal are divided into two parts, then the watermark bits are embedded by quantizing ratio of l p -norm of the two parts. Experimental results confirm the superiority of the proposed algorithm against common types attacks in comparison with the recently proposed scheme.

Non-stationary feature fusion of face and palmprint multimodal biometrics

Multimodal biometrics provides high performance in biometric recognition systems with respect to unibiometric systems as they offer a more universal approach, added security and better recognition accuracy. Moreover, data acquisition at the feature level brings out rich information from the traits, thus fusion of modalities at this level is desirable. In this paper we propose a novel fusion technique called non-stationary feature fusion where a new structure of interleaved matrix is constructed using local features extracted from two modalities i.e. face and palmprint images. A block based Discrete Cosine Transform (DCT) algorithm is used to construct a fused feature vector by extracting independent feature vectors from each spatial image. This fused feature vector contains nonlinear information that is used to train a Gaussian Mixture Models (GMM) based statistical model. The model provides accurate estimation of the class conditional probability density function of the fused feature vector. The proposed method produces recognition rates as high as 99.7% and 97% when tested on benchmark databases-ORL-PolyU and FERET-PolyU respectively. These rates are achieved using 23% low frequency DCT coefficients. The new technique is shown to outperform existing feature level fusion methods including methods based on matching and decision level fusion.

Watermarking‐based image authentication robust to JPEG compression

A watermarking-based image authentication scheme in the discrete cosine transform (DCT) domain robust to JPEG compression is presented. The binary authentication code is generated from a pseudo-random sequence based on a secret key and a block-dependent feature, protecting the scheme against cut-and-paste attacks. The watermark is embedded in low-frequency DCT coefficients selected by the secret key using a modified quantisation index modulation approach. Before embedding, the selected coefficients are quantised using the JPEG quantisation matrix for a selected quality factor, protecting the scheme against JPEG compression with higher quality factors. Experimental results show that the proposed technique achieves very good watermark imperceptibility and is able to detect and locate malicious attacks with good precision. Compared with other existing schemes, the proposed algorithm achieves better performance regarding false positive and false negative detection rates and in discriminating malicious attacks from JPEG compression.

Adaptive Image Watermarking Scheme Using Fuzzy Entropy and GA-ELM Hybridization in DCT Domain for Copyright Protection

A novel semi-blind image watermarking scheme based on fuzzy entropy and genetic algorithm (GA)--extreme learning machine (ELM) hybridization in discrete Cosine transform (DCT) domain for copyright protection is proposed in this paper. The selection of non overlapping blocks to embed the binary watermark is based on fuzzy entropy. As fuzzy entropy is able to discriminate data distribution under noise corrupted and redundant condition, feature extraction is more robust against various attacks. Each selected block is followed by 2-D DCT to transform it from spatial to frequency domain. Low frequency coefficients have good energy compactness and are robust to image processing attacks. As addition of noise corresponds to high frequency coefficients, these are not considered to embed the watermark in the proposed approach. The optimal scaling factor used to control the strength of watermark for each selected block of the image based on its noise sensitivity and tolerance limit is determined using GA optimization process. ELM is used to find an optimal regression function between the input feature vector (low frequency DCT coefficients) and corresponding target vector (in which the watermark bits are embedded) of each selected block. Then watermark embedding and extraction is performed intelligently by the regression function obtained by the trained ELM. The experimental results show that the proposed scheme is highly imperceptible and robust to geometric and non geometric attacks such as JPEG compression, filtering, noise addition, sharpening, gamma correction, scaling and cropping etc. To demonstrate the effectiveness of the proposed scheme, comparison with the state-of-art techniques clearly exhibits its applications for copyright protection.

An Improved Histogram-Based Features in Low-Frequency DCT Domain for Face Recognition

An Improved Histogram-Based Features in Low-Frequency DCT Domain for Face Recognition

The Afresh Transform Algorithm Based on Limited Histogram Equalization of Low Frequency DCT Coefficients

The Afresh Transform Algorithm Based on Limited Histogram Equalization of Low Frequency DCT Coefficients

Gray scale image watermarking using fuzzy entropy and Lagrangian support vector regression in DCT domain

A new image watermarking scheme based on fuzzy entropy and Lagrangian support vector regression (LSVR) in discrete cosine transform (DCT) domain is proposed in this paper. Fuzzy entropy is used to extract smooth non-overlapping blocks of an image followed by DCT to each selected block to transform it from spatial domain to frequency domain. Then a feature vector of low frequency DCT coefficients of each block act as an input to LSVR. The output (predicted value) of trained LSVR is used to embed the binary watermark by comparing it with the target value of the feature vector. As fuzzy entropy is able to discriminate data distribution under noise corrupted and redundant conditions, feature extraction is more robust against various attacks. The robustness of the proposed scheme is verified by performing various types of image processing operations which include geometric and non-geometric attacks. It is evident from experimental results that with the help of randomness measured of each image block using fuzzy entropy, good learning ability and high generalisation property of LSVR have made the proposed scheme more imperceptible and robust against different types of image processing operations.