International Journal of Pattern Recognition and Artificial Intelligence

Multi-label active learning for image classification has been a popular research topic. It faces several challenges, even though related work has made great progress. Existing studies on multi-label active learning do not pay attention to the cleanness of sample data. In reality, data are easily polluted by external influences that are likely to disturb the exploration of data space and have a negative effect on model training. Previous methods of label correlation mining, which are purely based on observed label distribution, are defective. Apart from neglecting noise influence, they also cannot acquire sufficient relevant information. In fact, they neglect inner relation mapping from example space to label space, which is an implicit way of modeling label relationships. To solve these issues, we develop a novel multi-label active learning with low-rank application (ENMAL) algorithm in this paper. A low-rank model is constructed to quantize noise level, and the example-label pairs that contain less noise are emphasized when sampling. A low-rank mapping matrix is learned to signify the mapping relation of a multi-label domain to capture a more comprehensive and reasonable label correlation. Integrating label correlation with uncertainty and considering sample noise, an efficient sampling strategy is developed. We extend ENMAL with automatic labeling (denoted as AL-ENMAL) to further reduce the annotation workload of active learning. Empirical research demonstrates the efficacy of our approaches.

Edge detection is an active and critical topic in the field of image processing, and plays a vital role for some important applications such as image segmentation, pattern classification, object tracking, etc. In this paper, an edge detection approach is proposed using local edge pattern descriptor which possesses multiscale and multiresolution property, and is named varied local edge pattern (VLEP) descriptor. This method contains the following steps: firstly, Gaussian filter is used to smooth the original image. Secondly, the edge strength values, which are used to calculate the edge gradient values and can be obtained by one or more groups of VLEPs. Then, weighted fusion idea is considered when multiple groups of VLEP descriptors are used. Finally, the appropriate threshold is set to perform binarization processing on the gradient version of the image. Experimental results show that the proposed edge detection method achieved better performance than other state-of-the-art edge detection methods.

Telecommunication traffic prediction is an important aspect of data analysis and processing in communication networks. In this study, we utilize the least-squares support vector machine (LSSVM) prediction method to improve the prediction performance of telecommunication traffic. As the parameters of LSSVM are difficult to determine, we propose to optimize the LSSVM parameters using the improved artificial bee colony (IABC) algorithm based on the fitness-prediction strategy (i.e. FP-IABC). We employ real traffic data collected on site to establish a telecommunication traffic forecasting model based on FP-IABC optimizing LSSVM (FP-IABC-LSSVM). The experiment results indicate that in the case involving no increase in the computational complexity, the proposed telecommunication traffic forecasting model-based FP-IABC-LSSVM has a higher prediction accuracy than the prediction model based on the ABC optimizing LSSVM (ABC-LSSVM), particle swarm optimizing LSSVM (PSO-LSSVM), and genetic algorithm optimizing LSSVM (GA-LSSVM). Further, with respect to the standard root mean square error and the average computation time, the proposed FP-IABC-LSSVM is the optimal prediction method of all of the comparison methods. The proposed prediction method not only improves the prediction accuracy, but also reduces the average computation time.

Overlapping community is a response to the real network structure in social networks and in real society in order to solve the problems such as the parameters of the existing overlapping community discovery algorithm being too large, excessive overlap and no guarantee of stability of multiple runs. In this paper, the method of calculating the node degree of membership was proposed, and an overlapping community discovery algorithm based on the local optimal expansion cohesion idea was designed. Firstly, the initial core community was constructed with the highest importance node and its neighbor nodes. Secondly, the core community was extended by node attribution degree until the termination condition of the algorithm was satisfied. Finally, the experimental results were compared with the existing algorithms. The experiments show that the result of the division by the improved algorithm has been significantly improved compared to the other algorithms, and the community structure after the division is more reasonable.

Recently, deep learning has been widely applied in various areas and achieved remarkable research findings. The major reason that makes the deep learning paradigm successful is that it can effectively learn a hierarchical feature structure for the training data. However, most deep learning algorithms rely on massive well-labeled training datasets and hyper-parameter configurations. This paper proposed a novel methodology that uses the geometric characteristics of line-segment representations to optimize the hyper-parameters for the deep networks. The methodology is applied to a line-segment-based stacked auto-encoder to verify its effectiveness. It is found that the line-segment-based visualizations can increase the interpretability of the deep models and facilitate the configurations for the hyper-parameters.

Sina Weibo, the most popular Chinese social platform with hundreds and millions of user-contributed images and texts, is growing rapidly. However, the noise between the image and text, as well as their incomplete correspondence, makes accurate image retrieval and ranking difficult. In this paper, we propose a deep learning framework using visual features, text content and popularity of Weibo to calculate the similarity between the image and the text based on training the model to maximize the likelihood of the target description sentence given the training image. In addition, the retrieval results are reranked using the popularity of the image. The comparison experiment of the large-scale Sina Weibo dataset proves the validity of the proposed method.

For image enhancement method based on the fractional order differential, it is difficult to artificially give the optimal order of the fractional differential which can make the image enhancement effect better, and it is hard to ensure the enhancement of the target object while preserving the information of background pixels if the entire image is filtered by a fixed differential order. In order to solve this problem, the image is segmented into the object area and the background area according to the Otsu threshold algorithm based on Markov Random Field firstly. On the basis of the principle of the fractional differential for image enhancement, a piecewise function is established by combining with the different characteristics of pixels in each area, then the best order of fractional differential in the two areas can be determined adaptively. Thus, a novel adaptive fractional order differential algorithm for image enhancement on the basis of segmentation is put forward. Several fog-degraded traffic images are selected for experiments and processed by three other algorithms. The results of comparison exhibit the superiority of our algorithm.

Screen content videos or images are widely used in applications such as screen sharing. Compressed screen content videos or images may have distortions or noises because of the quantization process. This paper proposes an improved sparse 3D transform-domain collaborative filter to enhance screen content image quality by block classification and block segmentation. Experimental results show that the proposed algorithm achieves a peak signal-to-noise ratio increase and subjective visual quality improvements for reconstructed screen content images.

This paper proposes a new oil painting reproduction method using 3D printing to compensate for the deficiencies of the existing methods. First, 3D reconstruction of oil paintings is completed by photogrammetry; the oil painting color and the 3D geometric information are recovered better by acquiring several sets of orthophotomaps, and modeling accuracy is ensured with a control mesh or by flattening. Next, the contours and hypsometric tints of the 3D model for oil paintings are generated using contour tracing algorithm, and the image segmentation of renderings is completed using RGB image segmentation algorithm, with the layered section extracted from each layer and the 3D geometric information converted into 2D plane information. Finally, the 3D models of oil paintings are presented through UV inkjet printing with images superimposed layer upon layer, and stereoscopic reproduction of oil paintings is completed based on the orthophotomaps printed from the 3D models.

This study designed an automatic sorting system that can separate haploid maize kernels from cross-breeding kernels that are marked with the Navajo label. This system comprises the seed feeding, image acquisition, sorting and system control units. The seed feeding unit distributes the maize kernels over the synchronous belt. The image acquisition unit acquires the images of maize kernels, as well as distinguishes the heterozygous from haploid kernels based on the color feature of the endosperm. The sorting unit contains mechanical arms and solenoid valves that can select the heterozygote kernels using air suction. Lastly, the system control unit is compatible with other units. Four maize varieties, namely, 958-CAU, 1050-37, LC0990-LN75 and LC0995-CAU, were provided by China Agricultural University, National Maize Improvement Center. These varieties were selected for the experiments conducted. The successful unloading rates of the system are 85.56%, 91.39%, 87.22% and 86.39%, respectively. (R$+$G)$\times |$G–B$|$/B was employed to distinguish the pigmented endosperm area. The accuracy rates of identification for the heterozygous kernels are 92.91%, 95.04%, 88.82% and 92.65% for 958-CAU, 1050-37, LC0990-LN75 and LC0995-CAU, respectively.

The video surveillance system based on face analysis has played an increasingly important role in the security industry. Compared with identification methods of other physical characteristics, face verification method is easy to be accepted by people. In the video surveillance scene, it is common to capture multiple faces belonging to a same person. We cannot get a good result of face recognition if we use all the images without considering image quality. In order to solve this problem, we propose a face deduplication system which is combined with face detection and face quality evaluation to obtain the highest quality face image of a person. The experimental results in this paper also show that our method can effectively detect the faces and select the high-quality face images, so as to improve the accuracy of face recognition.

In recent years, the fractional order derivative has been introduced for image enhancement. It was proved that the medical image enhancement method based on the fractional order derivative has better effect than the method based on the integral order calculus. However, a priori information such as texture surrounding a pixel is normally ignored by the traditional fractional differential operators with the same value in the eight directions. To address the above problem, this paper presents a new medical image enhancement method by taking the merits of fractional differential and directional derivative. The proposed method considers the surrounding information (such as the image edge, clarity and texture information) and structural features of different pixels, as well as the directional derivative of each pixel in constructing the masks. By proposing this method, it can not only improve the high frequency information, but also improve the low frequency information of the image. Ultimately, it enhances the texture information of the image. Extensive experiments on four kinds of medical image demonstrate that the proposed algorithm is in favor of preserving more texture details and superior to the existing fractional differential algorithms on medical image enhancement.

As we know, the nearest neighbor search is a good and effective method for good-sized image search. This paper mainly introduced how to learn an outstanding image feature representation form and a series of compact binary Hash coding functions under deep learning framework. Our concept is that binary codes can be obtained using a hidden layer to present some latent concepts dominating the class labels with usable data labels. Our method is effective in obtaining hash codes and image representations, so it is suitable for good-sized dataset. It is demonstrated in our experiment that the performances of the proposed algorithms were then verified on three different databases, MNIST, CIFAR-10 and Caltech-101. The experimental results reveal that two-proposed image Hash retrieval algorithm based on pixel-level automatic feature learning show higher search accuracy than the other algorithms; moreover, these two algorithms were proved to be more favorable in scalability and generality.

Multi-Instance Multi-Label learning (MIML) is a popular framework for supervised classification where an example is described by multiple instances and associated with multiple labels. Previous MIML approaches have focused on predicting labels for instances. The idea of tackling the problem is to identify its equivalence in the traditional supervised learning framework. Motivated by the recent advancement in deep learning, in this paper, we still consider the problem of predicting labels and attempt to model deep learning in MIML learning framework. The proposed approach enables us to train deep convolutional neural network with images from social networks where images are well labeled, even labeled with several labels or uncorrelated labels. Experiments on real-world datasets demonstrate the effectiveness of our proposed approach.