In pattern recognition and data mining, clustering is a classical technique to group matters of interest and has been widely employed to numerous applications. Among various clustering algorithms, K-means (KM) clustering is most popular for its simplicity and efficiency. However, with the rapid development of the social network, high-dimensional data are frequently generated, which poses a considerable challenge to the traditional KM clustering as the curse of dimensionality. In such scenarios, it is difficult to directly cluster such high-dimensional data that always contain redundant features and noises. Although the existing approaches try to solve this problem using joint subspace learning and KM clustering, there are still the following limitations: 1) the discriminative information in low-dimensional subspace is not well captured; 2) the intrinsic geometric information is seldom considered; and 3) the optimizing procedure of a discrete cluster indicator matrix is vulnerable to noises. In this paper, we propose a novel clustering model to cope with the above-mentioned challenges. Within the proposed model, discriminative information is adaptively explored by unifying local adaptive subspace learning and KM clustering. We extend the proposed model using a robust l2,1-norm loss function, where the robust cluster centroid can be calculated in a weighted iterative procedure. We also explore and discuss the relationships between the proposed algorithm and several related studies. Extensive experiments on kinds of benchmark data sets demonstrate the advantage of the proposed model compared with the state-of-the-art clustering approaches.

The accuracy of many computer vision tasks is reduced by blurred images, so deblur is important. More details of the image can be captured by a common multi-stage network, but the computational complexity of this method is higher compared with a single-stage network. However, a single-stage network cannot capture multi-scale information well. To tackle the problem, a novel convolutional encoder–decoder–restorer architecture is proposed. In this architecture, a multi-scale input structure is used in the encoder. Improved supervised attention module is inserted into the encoder for enhanced feature acquisition. In decoder, information supplement block is proposed to fuse multi-scale features. Finally, the fused features are used for image recovery in the restorer. In order to optimise the model in multiple domains, the loss function is calculated separately in the spatial and frequency domains. Our method is compared with existing methods on the GOPRO dataset. In addition, to verify the applications of our proposed method, we conduct experiments on the Real image dataset, the VOC2007 dataset and the LFW dataset. Experimental results show that our proposed method outperforms state-of-the-art deblurring methods and improves the accuracy of different vision tasks.

In the past few years, deep learning has become a research hotspot and has had a profound impact on computer vision. Deep CNN has been proven to be the most important and effective model for image processing, but due to the lack of training samples and huge number of learning parameters, it is easy to tend to overfit. In this work, we propose a new two-stage CNN image classification network, named “Improved Convolutional Neural Networks with Image Enhancement for Image Classification” and PLANET in abbreviation, which uses a new image data enhancement method called InnerMove to enhance images and augment the number of training samples. InnerMove is inspired by the “object movement” scene in computer vision and can improve the generalization ability of deep CNN models for image classification tasks. Sufficient experiment results show that PLANET utilizing InnerMove for image enhancement outperforms the comparative algorithms, and InnerMove has a more significant effect than the comparative data enhancement methods for image classification tasks.