Figure - available from: Signal Image and Video Processing
This content is subject to copyright. Terms and conditions apply.
Source publication
Increasing the quality of low-resolution images, namely super-resolution, has recently received a lot of attention in the field of image processing. Super-resolution has various applications, especially in the context of face recognition. Single-image super-resolution via sparse representation is the main issue of our proposed algorithm. A challeng...
Similar publications
Although previous studies have reported that individual prosocial traits generally enhance synchrony, gender as a social factor was not the subject of focus. The present study examined gender differences in synchrony by conducting three experiments on dyadic unstructured conversations with same‐gender strangers. Synchrony was assessed by calculatin...
The BOS method is a visualization method of density variation in the flow field. This method is simpler and cheaper to visualize than the Schlieren method. The primary purpose of this study is to propose WA-BOS (a simpler BOS method focusing on the amplitude of the Wavelet transform). Output images by WA-BOS and WP-BOS (BOS method focused on the ph...
Assessment of spatiotemporal characteristics of drought under climate change is significant for drought mitigation. In this study, the standardized precipitation evapotranspiration index (SPEI) calculated at different timescales was adopted to describe the drought conditions in the Heihe River Basin (HRB) from 1961 to 2014. The period characteristi...
![Figure (3): Process of wavelet transform [19]](profile/Israa-Muhsin-2/publication/335830396/figure/fig1/AS:803449755148292@1568568757300/Figure-3-Process-of-wavelet-transform-19_Q320.jpg)
Pan sharpening (fusion image) is the procedure of merging suitable information from two or more images into a single image. The image fusion techniques allow the combination of different information sources to improve the quality of image and increase its utility for a particular application. In this research, six pan-sharpening method have been im...
The conventional approach to pre-process data for compression is to apply transforms such as the Fourier, the Karhunen-Lo\`{e}ve, or wavelet transforms. One drawback from adopting such an approach is that it is independent of the use of the compressed data, which may induce significant optimality losses when measured in terms of final utility (inst...
Citations
... [34] introduced a discrete wavelet transform-based approach that uses sparse representation for creating a high-resolution image from a given low-resolution image. [35] introduced a new super-resolution technique for improving the steps of the sparse algorithm by allocating coefficients to specific patches in the dictionary. [36] developed a stacking learning-based approach for training the SR model. ...
The present article proposes a new Legendre wavelet (LW) filter-based image super-resolution technique. The LW basis vector is used in compact form using the unit step function. It is acquired using LW parameters and is discretized via collocation points. The LW basis is used to design a new filter that transforms the low-resolution image into a high-resolution image. This filter matrix is zoomed into a higher dimension using a discretized wavelet basis. The zero-padded low-resolution original image and the zoomed filter matrix are then multiplied by a constant parameter to enhance the information of the reconstructed image. The value of this constant parameter is optimized using the maximization of the entropy of the reconstructed image to further enhance the quality of the image. After the multiplication operation, the zoomed filter matrix and the zero-padded low-resolution original image are added to reconstruct the high-resolution image. Four existing techniques are compared with the proposed technique visually and through spatial frequency, standard deviation, and entropy. The proposed method shows better visual and quantitative feature performance as compared to other schemes.
... Sparse representation theory is widely used in face recognition [20], image denoising [21], target tracking and other fields because of its excellent data feature representation ability. Ma [22] established the target observation model with block orthogonal matching pursuit (BOMP) as the core, introduced the sparse display into the particle filtering framework, and found the optimal solution after optimization. ...
Fabric surface flaw inspection is essential for textile quality control, and it is demanding to replace human inspectors with the automatic machine vision-based flaw inspection system. To alleviate the time-consuming problem of sparse coding in detecting phase, this work presents a real-time fabric flaw inspection method by using grouped sparse dictionary. Firstly, the overcomplete sparse dictionary is learned from normal fabric images; secondly, the learned sparse dictionary is grouped into several sub-dictionaries by evaluating reconstruction error. Finally, the grouped dictionary is used to represent image and identify flaw regions as they cannot be represented well, leading to large reconstruction error. In addition, a non-maximum suppression algorithm is also proposed to reduce false inspection further. Experiments on various fabric flaws and real-time implementation on the proposed vision-based hardware system are conducted to evaluate the performance of proposed method. In comparison with other dictionary learning methods, the experimental results demonstrate that the proposed method can reduce the running time significantly and achieve a decent performance, which is capable of meeting the real-time inspection requirement without compromising inspection accuracy.
... In recent years, academics have focused more on developing Con-volutional Neural Network (CNN)based "SR" approaches. CNNs perform much better on "SR" tasks than classical machine learning models [19]. Several upgraded CNN models, inspired by the CNN-based SuperResolution Con-volutional NeuralNetwork "SRCNN" [20], haveebeen developed to attain better "SR"-performance. ...
High-resolution magnetic resonance imaging (MRI) benefits physicians in cancer detection and diagnosis, but such devices are hard to come by. Similarly, deep-learning-based image super-resolution technology may improve image quality. We describe a residual dense network "MRDN"-based technique to super-resolution MR image reconstruction for medicinal uses. We began by feeding the shallow layer's convolutional features to the residual dense block. The advantages were felt on both global and regional scales. Second, since each layer in the residual dense block is tightly coupled to the one below it, the features may be reused. Finally, we use a sub-pixel convolution layer for upsampling and super-resolution reconstruction to provide a sharp high-resolution image. Our "MRDN" technique surpasses traditional well-known super resolution algorithms in terms of PSNR and SSIM, two metrics of image quality. For quantitative testing, two public datasets are employed. The results of trials utilizing conventional evaluation criteria on these datasets show that the technique proposed in this study is superior. The resultant high-resolution MR image has a unique structure as well as several textural details.
... In recent years, academics have focused more on developing Con-volutional Neural Network (CNN)based "SR" approaches. CNNs perform much better on "SR" tasks than classical machine learning models [19]. Several upgraded CNN models, inspired by the CNN-based SuperResolution Con-volutional NeuralNetwork "SRCNN" [20], haveebeen developed to attain better "SR"-performance. ...
High-resolution magnetic resonance imaging (MRI) benefits physicians in cancer detection and diagnosis, but such devices are hard to come by. Similarly, deep-learning-based image super-resolution technology may improve image quality. We describe a residual dense network "MRDN"-based technique to super-resolution MR image reconstruction for medicinal uses. We began by feeding the shallow layer's convolutional features to the residual dense block. The advantages were felt on both global and regional scales. Second, since each layer in the residual dense block is tightly coupled to the one below it, the features may be reused. Finally, we use a sub-pixel convolution layer for upsampling and super-resolution reconstruction to provide a sharp high-resolution image. Our "MRDN" technique surpasses traditional well-known super resolution algorithms in terms of PSNR and SSIM, two metrics of image quality. For quantitative testing, two public datasets are employed. The results of trials utilizing conventional evaluation criteria on these datasets show that the technique proposed in this study is superior. The resultant high-resolution MR image has a unique structure as well as several textural details.
... Sparse representation theory is widely used in face recognition [20], image denoising [21], target tracking and other fields because of its excellent data feature representation ability. Ma [22] established the target observation model with BOMP as the core, introduced the sparse display into the particle filtering framework, and found the optimal solution after optimization. ...
Fabric surface flaw inspection is essential for textile quality control, and it is demanding to replace human inspectors with the automatic machine vision-based flaw inspection system. To alleviate the time-consuming problem of sparse coding in detecting phase, this work presents a real time fabric flaw inspection method by using grouped sparse dictionary. Firstly, the over-complete sparse dictionary is learned from normal fabric images; Secondly, the learned sparse dictionary is grouped into several sub-dictionaries by evaluating reconstruction error. Finally, the grouped dictionary is used to represent image and identify flaw regions as they cannot be represented well, leading to large reconstruction error. In addition, a non-maximum suppression algorithm is also proposed to reduce false inspection further. Experiments on various fabric flaws and real-time implementation on the proposed vision-based hardware system are conducted to evaluate the performance of proposed method. In comparison with other dictionary learning methods, the experimental results demonstrate that the proposed method can reduce the running time significantly and achieve a decent performance, which is capable of meeting the real-time inspection requirement without compromising inspection accuracy.
... It aims to extract the information from the original low-resolution (LR) image by algorithm, reconstruct the missing details, and then output the corresponding high-resolution (HR) image. This technology has important applications in many fields, including auxiliary diagnosis B Bin Chen chenbin121@swu.edu.cn 1 Chongqing [1][2][3], LR face recognition [4,5], and video quality improvement [6]. ...
Single image super-resolution (SISR) has important applications in many fields. With the help of this technology, the broadband requirement of image transmission can be reduced, the effect of remote sensing observation can be improved, and the location of lesion cells can be accurately located. Convolutional neural networks (CNNs) using multi-scale feature extraction structure can gain a large amount of information from a low-resolution input, which is helpful to improve the performance of SISR. However, these CNNs usually treat different types of information equally. There is a lot of redundancy in the information obtained, which limits the representation ability of the networks. We proposed an attention-enhanced multi-scale residual block (AMRB), which increases the proportion of useful information by embedding convolutional block attention module. Furthermore, we construct an attention-enhanced multi-scale residual network based on one time feature fusion (OAMRN). Extensive experiments illustrate the necessity of the AMRB and the superiority of proposed OAMRN over the state-of-the-art methods in terms of both quantitative metrics and visual quality.
... SISR establishes a mapping between LR image and HR image and uses the mapping to map a single LR image to the desired HR image. Traditional machine learning algorithms have been developed to learn the mapping, such as methods based on self-example learning [10], neighborhood embedding [11], and sparse representation [12]. Currently, the mainstream of SISR research is the methods based on deep learning. ...
Multiframe image super-resolution (MISR) combines complementary information of a set of low-resolution (LR) images to reconstruct a high-resolution (HR) one. In this study, we propose a robust and fully data-driven MISR method in the variational Bayesian framework. Different from the existing variational super-resolution (SR) methods, we use the l1 norm-based observation model, which takes the acquisition noise, outliers, and impulse noise into account. Furthermore, we have evaluated three typical image prior models, and the most appropriate one is chosen for our proposed method. The proposed method has the following advantages: (1) the HR image and all parameters are automatically estimated in an optimal stochastic sense; (2) the algorithm is robust to impulse noise and outliers. Extensive experiments with synthetic and real images demonstrate the advantages of the proposed method.
... Qiao et al. [22] proposed a regularization-based method for image interpolation where a new trainable reaction-diffusion model is used to interpolate an image. Fanaee et al. [8] proposed a new method to attain a high-resolution face image where discrete wavelet transform is combined with joint dictionary learning and sparse representation. Xi et al. [29] presented a super-resolution reconstruction algorithm that is based on deep self-encoding learning. ...
In this research, a new image enlargement scheme is proposed which is based on a hybrid combination of singular value decomposition (SVD) and the cubic spline interpolation method. The proposed scheme uses interpolation on SVD feature matrices which transfer the detailed information of lower dimension image to the higher dimension image in an efficient manner. Initially, the low-resolution image is split into two feature matrices and a weight matrix using SVD. These feature matrices are used in the cubic spline interpolation to find the intensity of intermediate pixels of the enlarged image. The step length of the interpolation function is optimized by the maximization of signal to noise ratio. The feature matrices obtained after the optimized interpolation are employed in the reconstruction of the output image. The proposed methodology is also applied to reconstruct the compressed images by taking a few most dominating eigenvectors of the interpolated images. The presented scheme is compared with three standard interpolation techniques visually and using entropy, signal to noise ratio, peak signal to noise ratio, and root mean square error as the performance measures. The results of the proposed work show superior visual performance and better SVD features transfer in the enlarged images as compared to the other schemes.
... In recent years, in order to take advantage of the sparsity and multiresolution of wavelet transform [29], a surge of approaches [30][31][32][33][34][35] with the wavelet technology have been proposed on image super resolution. Among these algorithms, [30][31][32][33] adopt the combination of the discrete wavelet transform and sparse representation instead of deep learning to obtain the HR image. ...
... In recent years, in order to take advantage of the sparsity and multiresolution of wavelet transform [29], a surge of approaches [30][31][32][33][34][35] with the wavelet technology have been proposed on image super resolution. Among these algorithms, [30][31][32][33] adopt the combination of the discrete wavelet transform and sparse representation instead of deep learning to obtain the HR image. Guo et al. [34] proposed DWSR as the first approach to predict high-resolution images in wavelet domain with a deep CNN network. ...
Medical imaging is widely used in medical diagnosis. The low-resolution image caused by high hardware cost and poor imaging technology leads to the loss of relevant features and even fine texture. Obtaining high-quality medical images plays an important role in disease diagnosis. A surge of deep learning approaches has recently demonstrated high-quality reconstruction for medical image super-resolution. In this work, we propose a light-weight wavelet frequency separation attention network for medical image super-resolution (WFSAN). WFSAN is designed with separated-path for wavelet sub-bands to predict the wavelet coefficients, considering that image data characteristics are different in the wavelet domain and spatial domain. In addition, different activation functions are selected to fit the coefficients. Inputs comprise approximate sub-bands and detail sub-bands of low-resolution wavelet coefficients. In the separated-path network, detail sub-bands, which have more sparsity, are trained to enhance high frequency information. An attention extension ghost block is designed to generate the features more efficiently. All results obtained from fusing layers are contracted to reconstruct the approximate and detail wavelet coefficients of the high-resolution image. In the end, the super-resolution results are generated by inverse wavelet transform. Experimental results show that WFSAN has competitive performance against state-of-the-art lightweight medical imaging methods in terms of quality and quantitative metrics.
... The continuous wavelet transform (CWT), which is based on wavelet analysis, can be also used as an alternative to STFT in order to obtain a TFR [22,23]. Wavelets are also used in image processing [24], engineering [25] and medicine [26,27]. The fidelity factor, Q, is the inverse of the relative bandwidth [28]. ...
... Therefore, the proposed transform could also be considered as multi-resolution STFT. In fact, if we denote STFT NW·Ts {x[m]}[m, f ] the STFT of a signal x[m], calculated with a fixed window size of NW · Ts seconds, the STFT-FD that we are proposing in this paper can be formulated by Eq. (24). ...
... Compared to wavelets, the selection of a different mother wavelet can lead to different results [25]. Although it can be considered a special case of multi-resolution STFT, that technique is typically computed processing only a couple of different values of window sizes, therefore adapting worse to different frequencies or scales [24]. The results of the STFT-FD were compared against standard STFT, an Adaptive Optimal-Kernel Time Frequency Representation, multi-resolution STFT and wavelets in [36,37] using several types of signals. ...
The short-time Fourier transform (STFT) is extensively used to convert signals from the time-domain into the time–frequency domain. However, the standard STFT has the drawback of having a fixed window size. Recently, we proposed a variant of that transform which fixes the window size in the frequency domain (STFT-FD). In this paper, we revisit that formulation, showing its similarity to existing techniques. Firstly, the formulation is revisited from the point of view of the STFT and some improvements are proposed. Secondly, the continuous wavelet transform (CWT) equation is used to formulate the transform in the continuous time using wavelet theory and to discretize it. Thirdly, the constant-Q transform (CQT) is analyzed showing the similarities in the equations of both transforms, and the differences in terms of how the sweep is carried out are discussed. Fourthly, the analogies with multi-resolution STFT are analyzed. Finally, the representations of a period chirp and an electrocardiogram signal in the time–frequency domain and the time-scale domain are obtained and used to compare the different techniques. The analysis in this paper shows that the proposed transform can be expressed as a variant of STFT, and as an alternative discretization of the CWT. It could also be considered a variant of the CQT and a special case of multi-resolution STFT.
