In this work, we generate Cayley graphs to obtain spectra corresponding to orthogonal eigenvectors that efficiently decorrelate multidimensional signals. Optimal signal decorrelation information is commonly obtained from the Karhunen-Loeve transform. However, whereas the traditional KL-transform autocovariance matrix is generated using multiple products between signal components, generating the Cayley graph only requires memory access operations for assigning weights to the graph edges. The graph is generated by coloring the edges between different vertices, where the vertices represent symmetry group elements, and the weight values are accessed from the discrete multidimensional functions. The efficiency from avoiding product calculations allows the use of spectral information for a wide variety of applications, as the sparsity of the spectra indicates the degree to which a function can be maximally decorrelated. For real-world applications, we begin by demonstrating how Cayley graph spectra can be used for one-dimensional signal classification. We extend our theory to multiple dimensions, allowing us to develop higher-dimensional applications. We obtain the Cayley graph representation of multiple two-dimensional image patches, which demonstrates that the spectra at each window patch can be used to match key points between transformed images. As another multidimensional example, we apply the method to a three-dimensional image patch; this is then used for boundary edge detection of three-dimensional scans. Thus, our method is useful for extracting valuable decorrelation information from signals of any number of dimensions, and it can outperform common signal processing techniques for a diverse set of tasks.
In this paper, we present a generalized framework of two-dimensional (2-D) Infinite impulse response (IIR) filters using offset binary coded distributed arithmetic technique (OBCDA). Based on the OBCDA formulations of 2-D IIR filter, several suitable multiplierless, reconfigurable realization architectures are proposed. Further, in this sharing of hardware based look up tables (HLUTs) in 2-D IIR filter is exploited to reduce the resources required by the proposed structures and pipelining is employed in some of the proposed structures to improve the throughput. For the realization of higher order multiplierless OBCDA based reconfigurable 2-D IIR filters, the internal building blocks of the filters is divided into smaller OBCDA blocks to reduce the increased complexity in the realization. Since, the proposed realization structure require lesser amount of resources (e.g. delay elements, adders, registers etc.), a substantial reduction in chip area and significant improvement in throughput and ADP can be observed. The ASIC synthesis is performed for various filter orders and comparative analysis is presented to show the effectiveness of our proposed realization structures over some of the previously reported structures. In particular, for filter order 3, one of our proposed 2-D IIR structures offers an improvement of 49.18
In this article, we present a spectral dependent depth Layering (SDDL) model designed to enhance the rendering quality of new views in light field rendering (LFR). The SDDL model adeptly utilizes minimal depth information within the scene, effectively mitigating spectral aliasing resulting from undersampling during multiview capture, thereby significantly improving the rendering quality of novel views. We conducted a detailed analysis of the phenomena and principles behind spectral aliasing in the light field and derived the precise scene depth information required to eliminate spectral aliasing. The quantity of depth layers employed quantitatively represents the depth information utilized within the scene, with a greater number of depth layers reducing the need for multiview captures. In instances of undersampling, the incidence of spectral aliasing is reduced. Building on the use of depth layers for spectral aliasing elimination and considering the variation in scene depth with the surface of an object, we propose a strategy that uses the surface curvature of the object to select pertinent depth information. By strategically selecting depth information, we were able to effectively eradicate spectral aliasing in the light field, optimizing the rendering quality of new views. We validated the spectral elimination efficacy of the SDDL method by capturing multiview data from both simulated scenes and real-world scenes. Furthermore, we tested the performance of the SDDL in enhancing the rendering quality of multiple views through comparative experiments. The experimental results demonstrate that the SDDL method achieves effective quality optimization performance.
Brain tumors are among the deadliest disease that affects the life of the human. One of the deadly diseases that affect the human population is the. It is a condition due to the damage that occurs in the chromosomes of particular cells or genes of the brain and due to this the brain cells halt its function. The techniques that are commonly utilized for the identification of brain tumors are Deep Learning, Computer-Aided Diagnosis, and Artificial Intelligence. Many of the existing techniques faced challenges in the accurate segmentation of various brain tumors. Hence, the 3D attention convolutional-based cross-modal distillation (3DAC-CMD) method is proposed to enhance automatic identification and MRI-based brain tumor segmentation. The datasets that are involved in the validation of the proposed method are Figshare Brain Tumor and Brain tumor MRI. Data preprocessing is carried out to facilitate training of the model by applying noise removal, normalization, contrast enhancement, image resizing, and image flipping. To transfer knowledge from one to other modalities, cross-model distillation is implemented. For a more precise estimation of brain tumor segmentation, a 3-dimensional Convolutional Neural Network is utilized. Attention Mechanism is employed to particularly focus on extracting information from area of interest. The existing methods utilized to compare the effectiveness of the proposed method are the AS-COA-E-U-Net, CJHBA, TISS-Net, RAAGR2-Net, and DenseUNet + . From the evaluations, it was found that the 3DAC-CMD method attained the highest performance metric values of 0.97 dice coefficient and 98.5
This system uses the Hessenberg decomposition (HD) and Chaotic structured phase masks (CSPMs) to propose a phase image encryption algorithm. HD makes the cryptosystem asymmetric. The use of CSPMs enhances the security of the proposed cryptosystem. The approach is further secured by the parameters of the Fresnel transform (FrT). The effectiveness of this method is examined using noise and occlusion attacks. To bolster its legitimacy, key sensitivity results were also included. The obtained values of mean squared error (MSE) is 1.68 × 10^ - 31 for grey image and 1.62 × 10^ - 31 for binary image which represents the effectiveness of the proposed cryptosystem. Parameters of Fresnel transform ( z_1 = 25 mm, z_2 = 35 mm, λ = 632.8 nm) are used to enhance the security of the proposed scheme. This outcome exhibits that the method is effective in obtaining safe image encryption, providing excellent image reconstruction with minimal information loss.
