International Journal of Image and Graphics

Parkinson’s disease (PD) is the most common disease that affects aged people which leads to dopamine-producing cells in substantia nigra to be damaged when motor system degenerates. Clinical Diagnosis of Parkinson’s disease at the earlier stage is very difficult. This work is carried out to find the significance of cognition function of basal ganglia (BG) region and speech data values. The BG can be segmented using morphological operation and active contour algorithm. Co-occurrences features are extracted and out of 720 features, the promising 110 features are selected using variance method. More promising 22 features are selected in speech data and both features are individually classified using SVM to find out the efficiency in Diagnosis. The outcome shows cognition function of BG performing a major role in early diagnosis of Parkinson’s disease when compared to speech data.

Segmentation is one of the critical steps in historical document image analysis systems that determines the quality of the search, understanding, recognition and interpretation processes. It allows isolating the objects to be considered and separating the regions of interest (paragraphs, lines, words and characters) from other entities (figures, graphs, tables, etc.). This stage follows the thresholding, which aims to improve the quality of the document and to extract its background from its foreground, also for detecting and correcting the skew that leads to redress the document. Here, a hybrid method is proposed in order to locate words and characters in both handwritten and printed documents. Numerical results prove the robustness and the high precision of our approach applied on old degraded document images over four common datasets, in which the pair (Recall, Precision) reaches approximately 97.7% and 97.9%.

In recent years, Image Registration has attracted lots of attention due to its capabilities and numerous applications. Various methods have been exploited to map two images with the same concept but different conditions. Considering the finding of the mentioned map as an optimization problem, mathematical-based optimization methods have been extensively employed due to their real-time performances. In this paper, we employed the Newton method to optimize two defined cost functions. These cost functions are Sum of Square Difference and Cross-Correlation. These presented algorithms have fast convergence and accurate features. Also, we propose an innovative treatment in order to attend to one of the free parameter-rotations or scale as a sole variable and the other one as the constant value. The assignment is replaced through the iterations for both parameters. The intuition is to turn a two-variable optimization problem into a single variable one in every step. Our simulation on benchmark images by the means of Root Mean Square Error and Mutual Information as the goodness criteria, that have been extensively used in similar studies, has shown the robustness and affectivity of the proposed method.

The satellite image classification plays a vital role in remote sensing for analyzing the images and recognizing the patterns. Supervised classification is one of the methods in which pixels of an image are grouped based on training samples. The uncertainty is one of the major issues in a supervised classification, where the pixel is classified into more than one class. This is happened due to the use of spectral values without considering contextual values in classification. Hence, this paper proposes Cellular Automata (CA)-based Classifier for Satellite Images Classification, where spectral values are combined with contextual values to improve the accuracy of the classifier. The proposed CA classifier combines the spectral values with contextual values in iteration until the uncertainty is resolved. Thereby, the proposed scheme improves the accuracy of the classical supervised classifier of parallel piped, minimum distance, DSVM and KNN classifier by 7% at an average.

Due to the technological advancements in digital communication, the amount of multimedia content over the internet has increased manifold in past decade. This has renewed the internet of researchers in the area of privacy and secure communication. This paper presents a secure and robust video steganography method in discrete cosine transform (DCT) domain. In order to enhance the security of the proposed algorithm, the frame selection process is randomized and the secret data are pre-treated using Arnold’s cat map. The secret data are embedded in the middle band DCT coefficient using two pseudo random sequences. These sequences are generated using a chaotic map. We analyze the proposed algorithm in terms of peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), multi-scale structural similarity index (MSSIM) and video quality metric (VQM). The evaluation has been done on 107 video sequences. The experimental results demonstrate that the algorithm maintains acceptable video quality. The robustness of the proposed method is tested under Gaussian and salt and pepper noise attack using correlation between original and recovered images. The proposed algorithm is able to recover 90.60% data without error under salt and pepper noise ($D=0.001$) attack and 87.23% data correctly under Gaussian noise attack with mean $=0$ and variance $=0.001$.

Machine learning techniques become more related to medical researches by using medical images as a dataset. It is categorized and analyzed for ultimate effectiveness in diagnosis or decision-making for diseases. Machine learning techniques have been exploited in numerous researches related to corneal diseases, contribution to ophthalmologists for diagnosing the diseases and comprehending the way automated learning techniques act. Nevertheless, confusion still exists in the type of data used, whether it is images, data extracted from images or clinical data, the course reliant on the type of device for obtaining them. In this study, the researches that used machine learning were reviewed and classified in terms of the kind of utilized machine for capturing data, along with the latest updates in sophisticated approaches for corneal disease diagnostic techniques.

Infrared and visible image fusion is a key area of research in multi-sensor image fusion. The main purpose of this fusion is to combine thermal information of the infrared image and texture information of the visible image. This paper presents an image fusion framework, based on parallel arrangement of sparse representation (SR) and spatial frequency (SF). In the proposed framework, an efficient edge-aware filter, i.e. guided filter, is first employed on the visible image. Then dual-tree complex wavelet transform (DTCWT) is used to obtain low-pass and high-pass coefficients of images, as it is shift-invariant and has high directional selectivity. The low-pass coefficients are fused using the SR- and SF-based fusion rules in parallel, which enhances the regional features of the images. The simulation results show that the proposed technique has better performance when compared with conventional techniques in both subjective and objective evaluations.

This paper proposes a multimodal biometric based authentication (verification and identification) with secured templates. Multimodal biometric systems provide improved authentication rate over unimodal systems at the cost of increased concern for memory requirement and template security. The proposed framework performs person authentication using face and fingerprint. Biometric templates are protected by hiding fingerprint into face at secret locations, through blind and key-based watermarking. Face features are extracted from approximation sub-band of Discrete Wavelet Transform, which reduces the overall working plane. The proposed method also shows high robustness of biometric templates against common channel attacks. Verification and identification performances are evaluated using two chimeric and one real multimodal dataset. The same systems, working with compressed templates provides considerable reduction in overall memory requirement with negligible loss of authentication accuracies. Thus, the proposed framework offers positive balance between authentication performance, template robustness and memory resource utilization.

Continuous monitoring and automatic detection of crowd activities is extremely helpful for management at public places to avoid any possible disaster. Analysis of crowded scene is a critical task as it typically involves poor resolution of objects, occlusions and complex dynamics. In this paper, we propose a novel, systematic and generalized method based on global motion analysis of people to detect Congestion situation in crowded scenes at entry/exit corridors. Our approach is tested on video footages acquired from surveillance cameras installed at exit corridors of public places. The results show the expediency of our approach.

Multifocus image fusion is a demanding research field due to the utilization of modern imaging devices. Generally, the scene to be captured contains objects at different distances from these devices and so a set of multifocus images of the scene is captured with different objects in-focus. However, to improve the situational awareness of the captured scene, these sets of images are required to be fused together. Therefore, a multifocus image fusion algorithm based on Convolutional Neural Network (CNN) and triangulated fuzzy filter is proposed. A CNN is used to extract information regarding focused pixels of input images and the same is used as fusion rule for fusing the input images. The focused information so extracted may still need to be refined near the boundaries. Therefore, asymmetrical triangular fuzzy filter with the median center (ATMED) is employed to correctly classify the pixels near the boundary. The advantage of using this filter is to rely on precise detection results since any misdetection may considerably degrade the fusion quality. The performance of the proposed algorithm is compared with the state-of-art image fusion algorithms, both subjectively and objectively. Various parameters such as edge strength ($Q$), fusion loss (FL), fusion artifacts (FA), entropy ($H$), standard deviation (SD), spatial frequency (SF), structural similarity index measure (SSIM) and feature similarity index measure (FSIM) are used to evaluate the performance of the proposed algorithm. Experimental results proved that the proposed fusion algorithm produces a fused image that contains all-in-one focused pixels and is better than those obtained using other popular and latest image fusion works.

A region of interest (ROI)-based compression method for medical image datasets is a requirement to maintain the quality of the diagnostically important region of the image. It is always a better option to compress the diagnostic important region in a lossless manner and the remaining portion of the image with a near-lossless compression method to achieve high compression efficiency without any compromise of quality. The predictive ROI-based compression on volumetric CT medical image is proposed in this paper; resolution-independent gradient edge detection (RIGED) and block adaptive arithmetic encoding (BAAE) are employed to ROI part for prediction and encoding that reduce the interpixel and coding redundancy. For the non-ROI portion, RIGED with an optimal threshold value, quantizer with optimal $q$-level and BAAE with optimal block size are utilized for compression. The volumetric 8-bit and 16-bit standard CT image dataset is utilized for the evaluation of the proposed technique, and results are validated on real-time CT images collected from the hospital. Performance of the proposed technique in terms of BPP outperforms existing techniques such as JPEG 2000, M-CALIC, JPEG-LS, CALIC and JP3D by 20.31%, 19.87%, 17.77%, 15.58% and 13.66%, respectively.

This paper proposes an image encryption technique which is fast and secure. The encryption scheme is designed for secure transmission of video surveillance data (keyframes) over insecure network. The image encryption technique employs 1D Sine–Sine system with better chaotic properties than its seed map and faster than higher-dimensional chaotic systems. Further, design of encryption scheme is based on two permutation rounds, which employs pixel swapping operation and diffusion operation which is simple and provides required security against plaintext, differential and various other attacks. Three separate chaotic sequences are generated using 1D Sine–Sine system which enhances the key space of the encryption scheme. Secret keys are updated dynamically with SHA-256 hash value obtained from plain image. Hash values of plain image are efficiently used without loss of any hash value information. This makes the encryption scheme plaintext sensitive and secure against plaintext attacks. Performance and security aspects of encryption scheme is analyzed both quantitatively using predefined security metrics and qualitatively by scrutinizing the internal working of encryption scheme. Computational complexity of encrypting a plain image of size $(rows\times columns)$ is ${𝒪}(rows\times columns)$ and is suitable for encrypting keyframes of video for secure surveillance applications.

Speckles are inherent to SAR. They hide and undermine several relevant information contained in the SAR images. In this paper, a despeckling algorithm using the shrinkage of two-dimensional discrete orthonormal S-transform (2D-DOST) coefficients in the transform domain along with shock filter is proposed. Also, an attempt has been made as a post-processing step to preserve the edges and other details while removing the speckle. The proposed strategy involves decomposing the SAR image into low and high-frequency components and processing them separately. A shock filter is used to smooth out the small variations in low-frequency components, and the high-frequency components are treated with a shrinkage of 2D-DOST coefficients. The edges, for enhancement, are detected using a ratio-based edge detection algorithm. The proposed method is tested, verified, and compared with some well-known models on C-band and X-band SAR images. A detailed experimental analysis is illustrated.

Sorting is a technique which is used to arrange the data in specific order. A sorting technique is applied to rearrange the elements in numerical order as ascending order or descending order or for words in alphabetical order. In this paper, we propose an efficient sorting algorithm known as Enhanced Bidirectional Insertion Sorting algorithm which is developed from insertion sort concept. A comparative analysis is done for the proposed Enhanced Bidirectional Insertion Sort algorithm with the selection sort and insertion sort algorithms. When compared to insertion sort algorithm the proposed algorithm outperforms with less number of comparisons in worst case and average case computing time. The proposed algorithm works efficiently for duplicated elements which is the advanced improvement and the results are proved.

The IT security paradigm evolves from secret-based to biometric identity-based. Biometric identification has gradually become more popular in recent years for handheld devices. Privacy-preserving is a key concern when biometrics is used in authentication systems in the present world today. Nowadays, the declaration of biometric traits has been imposed not only by the government but also by many private entities. There are no proper mechanisms and assurance that biometric traits will be kept safe by such entities. The encryption of biometric traits to avoid privacy attacks is a giant problem. Hence, state-of-the-art safety and security technological solutions must be devised to prevent the loss and misuse of such biometric traits. In this paper, we have identified different cancelable biometrics methods with the possible attacks on the biometric traits and directions on possible countermeasures in order to design a secure and privacy-preserving biometric authentication system. We also proposed a highly secure method for cancelable biometrics using a non-invertible function based on Discrete Cosine Transformation and Index of max hashing. We tested and evaluated the proposed novel method on a standard dataset and achieved good results.