Noise removal is a classic problem. Most researchers focus on Gaussian noise removal due to the regularity of the noise distribution, while mixed noise removal is always challenging because of the uncertainty of the noise distribution. Mixtures of additive white Gaussian noise (AWGN) with salt‐and‐pepper impulse noise (SPIN) and mixtures of AWGN with random‐valued impulse noise (RVIN) are typical examples of mixed noise. Most mixed noise removal methods are effective in the removal of mixed AWGN and SPIN, but perform poorly in the removal of AWGN and RVIN. The main reason is the randomness of RVIN, which leads to poor denoising performance when the RVIN is strong. In this paper, an improved nonlocal means‐based correction strategy (INS) is proposed. In INS, an improved nonlocal means strategy is applied to replace the impulse noise pixels to make the mixed noise obey an approximate Gaussian distribution. To prove the validity of INS, a convolutional neural network (CNN) in combination with INS (CNNINS) is applied to remove mixed noise. Experimental results are used to compare the proposed CNNINS with the most advanced mixed noise removal methods.

Polarimetric synthetic aperture radar (POLSAR) systems provide high resolution images containing polarimetric information. So, they have high capability in land cover classification. In this work, a binary coding‐based polarimetric‐morphological (BCPM) feature extraction is proposed for POLSAR image classification. At first, a set of polarimetric features is proposed. Then, a new morphological framework is introduced for contextual feature extraction from the POLSAR cube. The coherence matrix is composed from diagonal and non‐diagonal elements with different information. These elements are analysed separately in the proposed method. Moreover, the amplitude and phase components of the non‐diagonal elements are individually analysed using morphological filters by reconstruction. Finally, a binary coding‐based polarimetric‐spatial feature reduction, which uses the first order statistics, is proposed for feature transformation. The experiments on three real POLSAR images and a synthetic dataset show the superior performance of BCPM compared to several classification methods.

Coal‐rock fracture evolution has a key role in coal seam permeability. Due to the randomness and uncertainty of coal‐rock fractures, the prediction of fracture evolution is difficult and challenging. In this paper, the authors propose a dynamic‐difference based generative adversarial network (DDGAN) for coal‐rock fracture evolution prediction. Firstly, the spatial‐feature encoder and the dynamic‐difference encoder are proposed to capture the spatial features and the dynamic fracture evolution information independently. And a channel‐attention (CA) module is presented to enhance the contribution of fracture evolution details information in the dynamic‐difference encoder. Then, a multi‐scale fusion (MSF) module is proposed to fuse the spatial features and the dynamic‐difference features, which benefits to refine the detailed structure during decoding. Final, the compound objective function is employed to supervise and guide the network to achieve coal‐rock fracture evolution predictions. Compared with the state‐of‐the‐art methods, extensive experiment results demonstrate that the authors’ model can achieve better performance for the task of coal‐rock fracture evolution prediction.

Pixel‐based generative image inpainting has been widely researched over recent years and certain level of success via deep learning of feature representations and hallucinations of missing pixel values from surrounding backgrounds have also been reported in the literature. However, existing approaches rely on context‐based attentions and progressive inferences to capture the pixel correlations yet such pixel‐based approaches often fail to adapt to the constantly varying ranges and distances among surrounding background pixels. On the other hand, the modelling cost is also increasingly expensive whenever correlations of those pixels at longer distance away are to be exploited. To resolve the problem, we implement the principle of learning and hallucinating frequency components rather than pixel values. Therefore, we can avoid the dilemma that, on one hand the wish is to exploit all correlated pixels inside the image no matter how far away they are spatially located, but on the other, the price of increasing the modelling cost incurred by those pixels far away from the missing regions has to be paid. Extensive experiments carried out verify the effectiveness of the proposed method, which outperforms the representative existing state of the arts in terms of all assessment metrics.

Semantic segmentation is applied extensively in autonomous driving and intelligent transportation with methods that highly demand spatial and semantic information. Here, an STDC‐MA network is proposed to meet these demands. First, the STDC‐Seg structure is employed in STDC‐MA to ensure a lightweight and efficient structure. Subsequently, the feature alignment module is applied to understand the offset between high‐level and low‐level features, solving the problem of pixel offset related to upsampling on the high‐level feature map. The approach implements the effective fusion between high‐level features and low‐level features. A hierarchical multiscale attention mechanism is adopted to reveal the relationship among attention regions from two different input sizes of one image. Through this relationship, regions receiving much attention are integrated into the segmentation results, thereby reducing the unfocused regions of the input image and improving the effective utilisation of multiscale information. STDC‐MA maintains the segmentation speed as the STDC‐Seg network while improving the segmentation accuracy of small objects. STDC‐MA was verified on the validation dataset of Cityscapes. The segmentation result of STDC‐MA attained 78.32% mIOU with the input of 0.5× scale, 4.92% higher than STDC‐Seg.