The monocular depth estimation task has recently revealed encouraging prospects, especially for the autonomous driving task. To tackle the ill-posed problem of 3D geometric reasoning from 2D monocular images, multi-frame monocular methods are developed to leverage the perspective correlation information from sequential temporal frames. However, moving objects such as cars and trains usually violate the static scene assumption, leading to feature inconsistency deviation and misaligned cost values, which would mislead the optimization algorithm. In this work, we present CTA-Depth, a Context-aware Temporal Attention guided network for multi-frame monocular Depth estimation. Specifically, we first apply a multi-level attention enhancement module to integrate multi-level image features to obtain an initial depth and pose estimation. Then the proposed CTA-Refiner is adopted to alternatively optimize the depth and pose. During the CTA-Refiner process, context-aware temporal attention (CTA) is developed to capture the global temporal-context correlations to maintain the feature consistency and estimation integrity of moving objects. In particular, we propose a long-range geometry embedding (LGE) module to produce a long-range temporal geometry prior. Our approach achieves significant improvements (e.g., 13.5% for the Abs Rel metric on the KITTI dataset) over state-of-the-art approaches on three benchmark datasets.

Due to the lack of spatial cues giving merely a single image, many monocular depth estimation methods have been developed to leverage stereo or multi-view images to learn the spatial information of a scene in a self-supervised manner. However, these methods have limited performance gain since they are not able to exploit sufficient 3D geometry cues during inference, where only monocular images are available. In this work, we present V2Depth, a novel coarse-to-fine framework with Virtual View feature simulation for supervised monocular Depth estimation. Specifically, we first design a virtual-view feature simulator by leveraging the technique of novel view synthesis and contrastive learning to generate virtual view feature maps. In this way, we explicitly provide representative spatial geometry for subsequent depth estimation in both the training and inference stages. Then we introduce a 3DVA-Refiner to iteratively optimize the predicted depth map. During the optimization process, 3D-aware virtual attention is developed to capture the global spatial-context correlations to maintain the feature consistency of different views and estimation integrity of the 3D scene such as objects with occlusion relationships. Decisive improvements over state-of-the-art approaches on three benchmark datasets across all metrics demonstrate the superiority of our method.

In this paper, for efficient clustering of visual image data that have arbitrary mixture distributions, we propose a simple distance metric learning method called Maximum Normalized Spacing (MNS) which is a generalized principle based on Maximum Spacing [12] and Minimum Spanning Tree (MST). The proposed Normalized Spacing (NS) can be viewed as a kind of adaptive distance metric for contextual dissimilarity measure which takes into account the local distribution of the data vectors. Image clustering is a difficult task because there are multiple nonlinear manifolds embedded in the data space. Many of the existing clustering methods often fail to learn the whole structure of the multiple manifolds and they are usually not very effective. Combining both the internal and external statistics of clusters to capture the density structure of manifolds, MNS is capable of efficient and effective solving the clustering problem for the complex multi-manifold datasets in arbitrary metric spaces. We apply this MNS method into the practical problem of multi-view image clustering and obtain good results which are helpful for image browsing systems. Using the COIL-20 [19] and COIL-100 [18] multi-view image databases, our experimental results demonstrate the effectiveness of the proposed MNS clustering method and this clustering method is more efficient than the traditional clustering methods.

In recent years, many document image retrieval algorithms have been proposed. However, most of the current approaches either need good quality images or depend on the page layout structure. This paper presents a fast, accurate and OCR-free image retrieval algorithm using local feature sequences which can describe the intrinsic, unique and page-layout-free characteristics of document images. With a simple preprocessing step, the local feature sequences can be extracted without print-core detection and image registration. Then an efficient coarse-to-fine common substring matching strategy is applied to do local feature sequences matching. Beyond a single matching score, this approach can locate the matched parts word by word. It well handles the challenges including low resolution, different language, rotation and incompleteness and N-up. The encouraging experiment results on a large scale document image database show the retrieval outputs are sufficient good to be used directly as document image identification results.

In this paper, we present local patterns constrained image histograms (LPCIH) for efficient image retrieval. Extracting information through combining local texture patterns with global image histogram, LPCIH is an effective image feature representation method with a flexible image segmentation process. This kind of feature representation is robust and invariant for several image transforms, such as rotation, scaling and damaging. LPCIH method is efficient for several difficult image retrieval tasks, such as rotated and damaged gray image retrieval. Because many traditional image retrieval methods are not suitable for the difficult gray image retrieval tasks, LPCIH is valuable for many real-world applications of image retrieval. Experimental results show that the LPCIH method is consistently efficient, effective and it offers advantages over the state-of-the-art image retrieval methods.

We propose a discriminative feature selection method utilizing support vector machines for the challenging task of multiview face recognition. According to the statistical relationship between the two tasks, feature selection and multiclass classification, we integrate the two tasks into a single consistent framework and effectively realize the goal of discriminative feature selection. The classification process can be made faster without degrading the generalization performance through this discriminative feature selection method. On the UMIST multiview face database, our experiments show that this discriminative feature selection method can speed up the multiview face recognition process without degrading the correct rate and outperform the traditional kernel subspace methods.