Inferring all missing high-frequency details by using low-resolution image is the key to super-resolution reconstruction of a single image. In order to extract the feature information in lowresolution images fully and maximize the deduction of high-frequency details, we introduce a multi-scale cross merge (MSCM) network based on residual fusion. The MSCM uses feature extraction module with different-sized convolutional kernels to extract multiple features from the low resolution input image and send them into a nonlinear mapping module after concatenate them together. The nonlinear mapping module consists of five cross-merge modules, each of them formed by cascading three residual dual-branch merged structures. This structure can promote information integration of different branches. Dense connection and residual connection are integrated into the nonlinear mapping module to improve the transmission of information and gradient. The nonlinear mapping module is responsible for extracting highfrequency features and sending them to reconstruction module, which combines an improved sub-pixel upsampling layer with external residual and global residual to generate a high resolution image. Simulation experiments demonstrate that our MSCM network has the ability of achieving single-image superresolution reconstruction, and offers objective and subjective quality improvement compared to mainstream methods and other state-of-the-art reconstruction methods.

In order to achieve high-efficiency and high-precision multi-image classification tasks, a multi-attention ghost residual fusion network (MAGR) is proposed. MAGR is formed by cascading basic feature extraction network (BFE), ghost residual mapping network (GRM) and image classification network (IC). The BFE uses spatial and channel attention mechanisms to help the MAGR extract low-level features of the input image in a targeted manner. The GRM is formed by cascading 4 multi-branch group convolutional ghost residual blocks (MGR-Blocks). Each MGR-Block is cascaded by a dimension reducer and several ghost residual sub-networks (GRSs). The GRS integrates ghost convolution and residual connection, and the use of ghost convolution can significantly reduce parameters and achieve high-efficient classification. The GRS is a parallel convolution structure with 32 branches, which ensures that GRM has enough width to extract advanced features and extract as much feature information as possible, so as to obtain high-precision classification. The IC completes the aggregation of high-dimensional channel feature information, and then achieves a significant improvement in the classification accuracy of MAGR, by fusing the effective channel attention mechanism, global average pooling and SoftMax layer. Simulation experiment shows that MAGR has excellent classification capability while achieving high efficiency and lightweight. Compare with VGG16, the parameters of MAGR on CIFAR-10 is reduced by 94.8% while the classification accuracy is increased by 1.18%. Compare with MobileNetV2, the parameters of MAGR on CIFAR-100 is reduced by 33.9% while the classification accuracy is increased by 15.6%.

To protect edge and texture information, when removing salt‐and‐pepper (SP) noise in grayscale images, a support vector machine (SVM) denoising method is employed. First, a mapping relation between the neighborhood signal pixels and the central pixel is designed. The size of the neighborhood is a 5 × 5 region, with a signal pixel in the center. In this region, a 25‐dimensional input sample is constructed using the correlation between the neighborhood pixels and the eight‐direction fractional integral operators. The center signal pixel acts as the corresponding output sample to provide a training sample. Then, the SVM is trained with all training samples, and the SVM denoising model is obtained. Next, the center pixel value is estimated using the SVM denoising model in every 5 × 5 region with a noise pixel in the center. Finally, the noise pixel values are replaced with the estimated values of the SVM. The experiments demonstrate that the best denoising effect is obtained when the fractional integral order is in the range of 1.8 ± 0.1. The proposed method produces a visually pleasing denoised image and obtains superior image quality assessment indicators. Our method has significant advantages compared with state‐of‐the‐art denoisers when a low level of noise is present.

This paper focuses on the bearing estimation problem of far-field signal source via time-difference-of-arrival (TDOA) with a synchronized array in 3-D space. It is usually assumed that the propagation speed (PS) is perfectly known in localization. In reality, only an imperfect knowledge of PS could be obtained. The traditional closed-form solutions without involving PS have the advantage of low complexity, but suffer from low estimation accuracy. A measurement-division model is proposed to offer an alternative solution without the need of the propagation speed. This speed-free model combines two original TDOA measurement equations into a division formula, whose Cramer-Rao lower bound (CRLB) is derived for the observed data. A typical optimization method, i.e. the Levenberg-Marquardt (LM) algorithm is adopted to resolve the nonlinear measurement-division model, resulting in an estimation accuracy improvement because of its iterative search behavior. The theoretical performance of this solution is evaluated in terms of bias and covariance. Simulations are conducted to demonstrate an accuracy advantage of the solution over the related methods.