Multimedia Tools and Applications

We propose a novel single-image super resolution (SISR) approach using self-similarity of image and the low-rank matrix recovery (LRMR). The method performs multiple upsampling steps with relatively small magnification factors to recover a desired high resolution image. Each upsampling process includes the following steps: First, a set of low/high resolution (LR/HR) patch pairs is generated from the pyramid of the input low resolution image. Next, for each patch of the unknown HR images, similar HR patches are found from the set of LR/HR patch pairs by the corresponding LR patch and are stacked into a matrix with approximately low rank. Then, the LRMR technique is exploited to estimate the unknown HR image patch. Finally, the back-projection technique is used to perform the global reconstruction. We tested the proposed method on fifteen images including humans, animals, plants, text, and medical images. Experimental results demonstrate the effectiveness of the proposed method compared with several representative methods for SISR in terms of quantitative metrics and visual effect.

Digital media is often handled in a compressed and encrypted form in Digital Asset Management Systems. And watermarking of the compressed encrypted media items in the compressed-encrypted domain itself is required sometimes for copyright violation detection or other purposes. In this paper, we propose a robust image watermarking technique for partially compressed-encrypted JPEG images. However, arbitrary embedding of a watermark in a partially compressed encrypted image can cause drastic degradation of the quality as the underlying change may result in random decrypted values. In addition, due to the encryption the compression efficiency may become very low. Thus the challenge is to design a watermarking technique that provides good watermarked image quality and at the same time gives good compression efficiency. While the proposed technique embeds watermark in the partially compressed-encrypted domain, the extraction of watermark can be done in the encrypted or decrypted domains. The experiments show that the watermarked image quality is good and the reduction in compression efficiency is low. The proposed watermarking technique is robust to common signal processing attacks. The watermark detection performance of the proposed scheme is better than the existing encrypted domain watermarking techniques.

The performance of video surveillance systems with network cameras depends on their accuracy in people re-identification. Body occlusion, crowded background, and variations in scene illumination and pose are challenging issues in people re-identification. In this paper, a technique is proposed to improve the performance of re-identification approaches using (a) a pre-processing step; and (b) a proposed weighing mechanism. In this approach, first, the input image is segmented into the person’s body, background, and possible carried objects. Then, considering the image’s segments, the occluded parts of the body are retrieved using their neighboring pixels. The processed image is transformed into the log chromatically color space which is robust to scene illumination changes. Using the transformed images along with descriptors which are robust to appearance changes such as Gaussian of Gaussian (GOG) and Hierarchical Gaussian Descriptor (HGD) can improve performance of the descriptors. In this paper, the GOG and HGD are used in a weighed form to represent the pre-processed images considering the importance of each segment of the images in people re-identification. The proposed re-identification system is evaluated using VIPeR and PRID450s datasets, where it respectively achieves 61.9% and 83.4% rank-1 matching rates. Experimental results show that our proposed approach outperforms other existing approaches in people re-identification.

The fuzzy linking color histogram considers not only the similarity of different colors from different bins but also the dissimilarity of those colors assigned to the same bin. Moreover, it projects the three-dimension histogram onto the one single-dimension histogram, which reduces the complexity of computation. Spatial fuzzy linking color histogram (SFLCH) combines fuzzy linking color histogram with spatial information that describes the color distribution of pixels in different regions. Meanwhile, the concept “color complexity” is defined in histogram similarity measure in order to add the influence of human vision perception to image retrieval. Compared with other methods, the modified fuzzy color histogram is proved to be more accurate and effective for the content-based image retrieval from the experimental results.

Information hiding researchers have been exploring techniques to improve the security of transmitting sensitive data through an unsecured channel. This paper proposes an audio steganography model for secure audio transmission during communication based on fractal coding and a chaotic least significant bit or also known as HASFC. This model contributes to enhancing the hiding capacity and preserving the statistical transparency and security. The HASFC model manages to embed secret audio into a cover audio with the same size. In order to achieve this result, fractal coding is adopted which produces high compression ratio with the acceptable reconstructed signal. The chaotic map is used to randomly select the cover samples for embedding and its initial parameters are utilized as a secret key to enhancing the security of the proposed model. Unlike the existing audio steganography schemes, The HASFC model outperforms related studies by improving the hiding capacity up to 30% and maintaining the transparency of stego audio with average values of SNR at 70.4, PRD at 0.0002 and SDG at 4.7. Moreover, the model also shows resistance against brute-force attack and statistical analysis.

In this article is introduced an innovative segmentation methodology that is based on a two-dimensional (2D) histogram that permits to increase the quality of the segmented images. The 2D histogram is constructed using the Lateral Inhibition (LI) that helps maintain and remark different image features. To segment the image in the proposed approach, the 2D Rényi entropy is used, which is a multi-level thresholding technique. Since the complexity of the 2D Rényi entropy increases with the number of thresholds, it is necessary to use an efficient search mechanism. To perform this task, it is also proposed an improved version of the Electromagnetic Field Optimization (EFO) algorithm that employs the High Disruptive Polynomial Mutation (HDPM) to exploit the search space intensively. The proposed metaheuristic is called MEFO. In combination with the 2D Rényi entropy creates a robust mechanism able to find the optimal configuration of thresholds that permits an accurate classification of the information contained in the 2D histogram generated using the (LI). The performance of the MEFO is tested over the Berkeley Segmentation Dataset (BSDS100) that contains 100 images with different complexities. The experiments include quantitative, qualitative, and statistical tests that permit the MEFO's efficiency in both senses for image segmentation and for solving multidimensional real optimization problems. Moreover, different comparisons validate the capabilities of the proposed algorithms to segment the images properly.

Osteoarthritis (OA) is a common form of knee arthritis which causes significant disability and is threatening to plague patient’s quality of life. Although this chronic condition does not lead to fatality, still there exists no known cure for OA. Diagnosis of OA can be confirmed primarily based on radiographic findings. Being a progressive disease, early identification of OA is crucial for clinical interventions to curtail the OA degeneration. Kellgren-Lawrence (KL) grading system has been traditionally employed to assess the knee OA severity. Due to the recent advancements of deep learning in computer vision, more studies have employed deep neural network in automatically predicting KL grade from plain knee joint radiograph. However, these studies treat KL grading as a multi-class classification task and ignore the inherent ordinal nature within the KL grades. In this study, we propose an ordinal regression module for neural networks to treat KL grading as an ordinal regression task. Our module takes an input from neural network and produces 4 cut-points to partition the prediction space into 5 respective KL grades. The proposed model is optimized by a cumulative-link loss function. Performance of the model is evaluated against various notable neural networks and significant improvements on the knee OA KL grade prediction were demonstrated.