Explore the vast range of titles available.

In recent steganographic literature, video steganography becomes popular due to its capability of accommodating higher payload. Since the video is transmitted mostly in a compressed format, compressed domain parameters are a natural choice for data embedding. In this paper, a motion vector based video steganographic method is proposed. For embedding the secret bit stream, the embedding motion vectors are selected for the homogeneous regions of the reference frame. Since homogeneous or smooth regions contain macro blocks with similar prediction error blocks, it helps to reduce the chance of detection by masking the embedding noise with similar prediction error among neighbouring macro blocks. The efficient search window and polar orientation based embedding technique are used to improve the imperceptibility against standard steganalysis schemes. A set of experiments is been carried out to justify the efficacy of the proposed scheme over the related existing steganographic methods.

Steganalysis is the science for detecting steganographic traces in innocent-looking digital media like images, videos, etc. In recent literature, it has been observed that state-of-the-art image steganographic techniques such as S-UNIWARD, HUGO, WOW, etc. still remain undetected even with considerable embedding payload. Recently, the deep learning framework has been hugely successful in different computer vision applications like object detection, image classification, event detection, etc. Some recent deep learning-based works also show promising results for image steganalysis and have opened a new avenue for research. The current literature reveals that the steganalytic detector becomes more precise if trained on the residual error (embedding noise) domain. To get an accurate noise residual, it is required to predict the cover image precisely from the corresponding stego image. In this work, a denoising kernel has been learned to obtain a more precise noise residual. After that, a CNN based steganalytic detector is devised, which is trained using the noise residual to get a more precise detection. Experimental results show that the proposed scheme outperforms the state-of-the-art steganalysis schemes against the state-of-the-art steganographic approaches.

This paper proposes a novel compressed domain robust watermarking scheme which embeds watermark by altering the intra prediction modes of 4 × 4 intra prediction blocks of the most recent high-definition video standards H.265/HEVC. Due to different compression architecture and higher number of intra prediction mode, the existing intra prediction mode based watermarking strategies for previous video standards such as H.264/AVC are not robust when those are directly applied for H.265/HEVC. This proposed work overcomes this shortcoming by reducing the synchronization error of watermark after re-compression attack in two stage. First, a spatial texture analysis is done based on number of non-zero transform coefficients of embedding blocks. Then, suitable candidate blocks for watermark embedding are selected based on 4 × 4 intra luma PB’s sustainably and watermarked mode sustainability while maintaining visual quality and bit rate. In next stage, the robustness of the proposed method has been enhanced by grouping of intra prediction modes such a way that mode change due to re-compression can be closed within a group. Finally, each group is represented with two bits of watermark sequence and embedding have been done by altering prediction modes of selected 4 × 4 intra prediction block to the representative mode of the group denoted by the watermark bit pair. Experimental results on various test sequences show that the scheme is robust against re-compression with high QP values and robustness has been increased twice compared to existing intra prediction mode based watermarking schemes. Also, the proposed scheme has very low effect on the visual quality having least peak to signal ration of 28 dB for the watermarked test sequences and also has very similar bit increase rate compared to existing scheme.

High-quality image de-raining is a challenging task that has been given considerable importance in recent times. To begin with, this problem is modeled as an image decomposition task where a rainy image is decomposed into the rain-free background and the associated rain streak map. Most of the existing methods have been successful in removing the rain-streaks but fails to restore the image quality, which is degraded due to noise removal. This paper proposes a novel architecture called High-Resolution Image De-Raining using Conditional Generative Adversarial Networks (HRID-GAN) to generate a de-rained image with minimal artifacts and better visual quality. Extensive experiments on publicly available synthetic as well as real-world datasets show a substantial improvement over the state-of-the-art methods SPANet (Wang et al. 2019) by ∼ 2.43% in PSNR and, DID-MDN (Zhang and Patel 2018) by ∼ 2.43%, ∼ 10.12% and ID-CGAN (Zhang et al. 2017) by ∼ 11.80%, ∼ 34.70% in SSIM and PSNR respectively.

In recent time, most of the steganographic methods minimize the embedding cost while maximizing the embedding capacity by injecting message bits in the highly textured regions of the image. Recently, the Clustering Modification Direction (CMD) steganography has been proposed as a wrapper over the additive steganography algorithms, resulting in a substantial improvement in statistical imperceptibility against state-of-the-art steganalytic classifiers. In this paper, a steganalysis scheme, named Selective-Signal-Removal (SSR) is proposed to mount an attack on the CMD algorithm. It has been observed experimentally that the CMD scheme has a tendency to embed in a localized cluster having higher texture. The proposed scheme exploits this fact and tries to predict the embedding zones. It essentially discards the irrelevant region of the image (which may not be modified by the CMD algorithm while embedding) by using a heuristic function with an assignment algorithm to improve the steganalytic detection rate. The experimental results show that the proposed SSR scheme can detect CMD based steganography with improved accuracy.

Image memorability is a recent topic in the domain of computer vision, which enables one to measure the degree at which images are memorable to human cognitive system. Initial research on image memorability shown that memorability is an inherent characteristic of an image, and humans are consistent in remembering images. Further, it is also demonstrated that memorability of an image can be determined using machine learning and computer vision techniques. In this paper, a novel deep learning based image memorability prediction model is proposed. The proposed model automatically learns and utilises multiple visual factors such as object semantics, visual emotions, and saliency to predict image memorability scores. In particular, the proposed model employs multiple instance learning framework to utilise emotion cues evoking from single global context and multiple local contexts of an image. An extensive set of experiments are being carried out on large-scale image memorability dataset LaMem. The experimental results show that the proposed model performs better than current state-of-the-art models by reaching a rank correlation of 0.67, which is close to human consistency (ρ = 0.68).

In recent times, deep learning-based steganalysis classifiers have become popular due to their state-of-the-art performance. Most deep steganalysis classifiers usually extract noise residuals using high-pass filters as preprocessing steps and feed them to their deep model for classification. It is observed that recent steganographic embedding does not always restrict their embedding in the high-frequency zone; instead, they distribute it as per embedding policy. Therefore, besides noise residual, learning the embedding zone is another challenging task. In this work, unlike the conventional approaches, the proposed model first extracts the noise residual using learned denoising kernels to boost the signal-to-noise ratio. After preprocessing, the sparse noise residuals are fed to a novel Multi-Contextual Convolutional Neural Network (M-CNET) that uses heterogeneous context size to learn the sparse and low-amplitude representation of noise residuals. The model performance is further improved by incorporating the Self-Attention module to focus on the areas prone to steganalytic embedding. A set of comprehensive experiments is performed to show the proposed scheme’s efficacy over the prior arts. Besides, an ablation study is given to justify the contribution of various modules of the proposed architecture.

In 3D image compression, depth image based rendering (DIBR) is one of the latest techniques where the center image (say the main view, is used to synthesise the left and the right view image) and the depth image are communicated to the receiver side. It has been observed in the literature that most of the existing 3D image watermarking schemes are not resilient to the view synthesis process used in the DIBR technique. In this paper, a 3D image watermarking scheme is proposed which is invariant to the DIBR view synthesis process. In this proposed scheme, 2D-dual-tree complex wavelet transform (2D-DT-CWT) coefficients of centre view are used for watermark embedding such that shift invariance and directional property of the DT-CWT can be exploited to make the scheme robust against view synthesis process. A comprehensive set of experiments has been carried out to justify the robustness of the proposed scheme over the related existing schemes with respect to the JPEG compression and synthesis view attack.

Single image rain streak removal is a well-explored topic in the field of computer vision. The de-raining problem is modeled as an image decomposition task where a rainy image is decomposed into rain-free background image and rain streek map. Unlike most of the existing de-raining methods, this paper attempts to decompose the rainy image in the frequency domain. The idea is inspired by pseudo-periodic characteristics of the noise signal (here the rain streaks) which leave some traces in the frequency domain, and the same can be utilized to predict the noise signal. In this paper, a deep learning-based rain streak prediction model is proposed which learns in discrete Fourier transform Oppenheim and Schafer (Discrete-TimeSignal Processing, Prentice Hall, Upper Saddle River, 1989) domain. To the best of our knowledge, this is the first approach where compressed domain coefficients are directly used as input to a deep convolutional neural network. The proposed model has been tested on publicly available synthetic datasets Fu et al. (in: 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017. https://doi.org/10.1109/CVPR.2017.186 , Yang et al. (in: 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017. https://doi.org/10.1109/CVPR.2017.183 ), Yeh et al. (in: 2015 IEEE International Conference on Consumer Electronics-Taiwan, 2015. https://doi.org/10.1109/ICCE-TW.2015.7216999 ) and results are found to be comparable with the state of the art methods in the spatial domain. The presented analysis and study have an obvious indication to extend transform domain input to train the deep learning architecture especially image de-noising like problems.

Knee Osteoarthritis (OA) is a destructive joint disease identified by joint stiffness, pain, and functional disability concerning millions of lives across the globe. It is generally assessed by evaluating physical symptoms, medical history, and other joint screening tests like radiographs, Magnetic Resonance Imaging (MRI), and Computed Tomography (CT) scans. Unfortunately, the conventional methods are very subjective, which forms a barrier in detecting the disease progression at an early stage. This paper presents a deep learning-based framework, namely OsteoHRNet, that automatically assesses the Knee OA severity in terms of Kellgren and Lawrence (KL) grade classification from X-rays. As a primary novelty, the proposed approach is built upon one of the most recent deep models, called the High-Resolution Network (HRNet), to capture the multi-scale features of knee X-rays. In addition, an attention mechanism has been incorporated to filter out the counterproductive features and boost the performance further. Our proposed model has achieved the best multi-class accuracy of 71.74% and MAE of 0.311 on the baseline cohort of the OAI dataset, which is a remarkable gain over the existing best-published works. Additionally, Gradient-based Class Activation Maps (Grad-CAMs) have been employed to justify the proposed network learning.