CVPR 2015 论文列表

Photographs taken through glass windows often contain both the desired scene and undesired reflections. Separating the reflection and transmission layers is an important but ill-posed problem that has both aesthetic and practical applications. In this work, we introduce the use of ghosting cues that introduce asymmetry between the layers, thereby helping to significantly reduce the ill-posedness of the problem. Such cues arise from shifted double reflections of the reflected scene off the glass surface. In double-pane windows, each pane reflects shifted and attenuated versions of objects on the same side of the glass as the camera. For single-pane windows, ghosting cues arise from shifted reflections on the two surfaces of the glass pane. Even though the ghosting is sometimes barely perceptible by humans, we can still exploit the cue for layer separation. In this work, we model the ghosted reflection using a double-impulse convolution kernel, and automatically estimate the spatial separation and relative attenuation of the ghosted reflection components. To separate the layers, we propose an algorithm that uses a Gaussian Mixture Model for regularization. Our method is automatic and requires only a single input image. We demonstrate that our approach removes a large fraction of reflections on both synthetic and real-world inputs.

We introduce a novel method for using reflectance to identify materials. Reflectance offers a unique signature of the material but is challenging to measure and use for recognizing materials due to its high-dimensionality. In this work, one-shot reflectance of a material surface which we refer to as a reflectance disk is capturing using a unique optical camera. The pixel coordinates of these reflectance disks correspond to the surface viewing angles. The reflectance has class-specific stucture and angular gradients computed in this reflectance space reveal the material class. These reflectance disks encode discriminative information for efficient and accurate material recognition. We introduce a framework called reflectance hashing that models the reflectance disks with dictionary learning and binary hashing. We demonstrate the effectiveness of reflectance hashing for material recognition with a number of real-world materials.

In recent years, the convolutional neural network (CNN) has achieved great success in many computer vision tasks. Partially inspired by neuroscience, CNN shares many properties with the visual system of the brain. A prominent difference is that CNN is typically a feed-forward architecture while in the visual system recurrent connections are abundant. Inspired by this fact, we propose a recurrent CNN (RCNN) for object recognition by incorporating recurrent connections into each convolutional layer. Though the input is static, the activities of RCNN units evolve over time so that the activity of each unit is modulated by the activities of its neighboring units. This property enhances the ability of the model to integrate the context information, which is important for object recognition. Like other recurrent neural networks, unfolding the RCNN through time can result in an arbitrarily deep network with a fixed number of parameters. Furthermore, the unfolded network has multiple paths, which can facilitate the learning process. The model is tested on four benchmark object recognition datasets: CIFAR-10, CIFAR-100, MNIST and SVHN. With fewer trainable parameters, RCNN outperforms the state-of-the-art models on all of these datasets. Increasing the number of parameters leads to even better performance. These results demonstrate the advantage of the recurrent structure over purely feed-forward structure for object recognition.

This paper describes a method for recovering appearance of inner slices of translucent objects. The outer appearance of translucent objects is a summation of the appearance of slices at all depths, where each slice is blurred by depth-dependent point spread functions (PSFs). By exploiting the difference of low-pass characteristics of depth-dependent PSFs, we develop a multi-frequency illumination method for obtaining the appearance of individual inner slices using a coaxial projector-camera setup. Specifically, by measuring the target object with varying the spatial frequency of checker patterns emitted from a projector, our method recovers inner slices via a simple linear solution method. We quantitatively evaluate accuracy of the proposed method by simulations and show qualitative recovery results using real-world scenes.

We propose a novel, large-scale, structure-from-motion framework that advances the state of the art in data scalability from city-scale modeling (millions of images) to world-scale modeling (several tens of millions of images) using just a single computer. The main enabling technology is the use of a streaming-based framework for connected component discovery. Moreover, our system employs an adaptive, online, iconic image clustering approach based on an augmented bag-of-words representation, in order to balance the goals of registration, comprehensiveness, and data compactness. We demonstrate our proposal by operating on a recent publicly available 100 million image crowd-sourced photo collection containing images geographically distributed throughout the entire world. Results illustrate that our streaming-based approach does not compromise model completeness, but achieves unprecedented levels of efficiency and scalability.

A considerable portion of web images capture events that occur in our personal lives or social activities. In this paper, we aim to develop an effective method for recognizing events from such images. Despite the sheer amount of study on event recognition, most existing methods rely on videos and are not directly applicable to this task. Generally, events are complex phenomena that involve interactions among people and objects, and therefore analysis of event photos requires techniques that can go beyond recognizing individual objects and carry out joint reasoning based on evidences of multiple aspects. Inspired by the recent success of deep learning, we formulate a multi-layer framework to tackle this problem, which takes into account both visual appearance and the interactions among humans and objects, and combines them via semantic fusion. An important issue arising here is that humans and objects discovered by detectors are in the form of bounding boxes, and there is no straightforward way to represent their interactions and incorporate them with a deep network. We address this using a novel strategy that projects the detected instances onto multi-scale spatial maps. On a large dataset with $60,000$ images, the proposed method achieved substantial improvement over the state-of-the-art, raising the accuracy of event recognition by over $10\%$.

Microcirculatory monitoring plays an important role in diagnosis and treatment of critical care patients. Sidestream Dark Field (SDF) imaging devices have been used to visualize and support interpretation of the micro-vascular blood flow. However, due to subsurface scattering within the tissue that embeds the capillaries, transparency of plasma, imaging noise and lack of features, it is difficult to obtain reliable physiological data from SDF videos. Therefore, thus far microcirculatory videos have been analyzed manually with significant input from expert clinicians. In this paper, we present a framework that automates the analysis process. It includes stages of video stabilization, enhancement, and micro-vessel extraction, in order to automatically estimate statistics of the micro blood flows from SDF videos. Our method has been validated in critical care experiments conducted carefully to record the microcirculatory blood flow in test animal subjects before, during and after induced bleeding episodes, as well as to study the effect of fluid resuscitation. Our method is able to extract microcirculatory measurements that are consistent with clinical intuition and it has a potential to become a useful tool in critical care medicine.

Robust object tracking is a challenging task in computer vision. To better solve the partial occlusion issue, part-based methods are widely used in visual object trackers. However, due to the complicated online training and updating process, most of these part-based trackers cannot run in real-time. Correlation filters have been used in tracking tasks recently because of the high efficiency. However, the conventional correlation filter based trackers cannot deal with occlusion. Furthermore, most correlation filter based trackers fix the scale and rotation of the target which makes the trackers unreliable in long-term tracking tasks. In this paper, we propose a novel tracking method which track objects based on parts with multiple correlation filters. Our method can run in real-time. Additionally, the Bayesian inference framework and a structural constraint mask are adopted to enable our tracker to be robust to various appearance changes. Extensive experiments have been done to prove the effectiveness of our method.

The widely-used approach for estimating camera orientation is to use points at infinity, i.e., vanishing points (VPs). By enforcing the orthogonal constraint between the VPs, called the Manhattan world constraint, a drift-free camera orientation estimation can be achieved. However, in practical applications this approach suffers from many spurious parallel line segments or does not perform in non-Manhattan world scenes. To overcome these limitations, we propose a novel method that jointly estimates the VPs and camera orientation based on sequential Bayesian filtering. The proposed method does not require the Manhattan world assumption, and can perform a highly accurate estimation of camera orientation in real time. In addition, in order to enhance the robustness of the joint estimation, we propose a feature management technique that removes false positives of line clusters and classifies newly detected lines. We demonstrate the superiority of the proposed method through an extensive evaluation using synthetic and real datasets and comparison with other state-of-the-art methods.

In this paper, we tackle the problem of unsupervised segmentation in the form of superpixels. Our main emphasis is on speed and accuracy. We build on [31] to define the problem as a boundary and topology preserving Markov random field. We propose a coarse to fine optimization technique that speeds up inference in terms of the number of updates by an order of magnitude. Our approach is shown to outperform [31] while employing a single iteration. We evaluate and compare our approach to state-of-the-art superpixel algorithms on the BSD and KITTI benchmarks. Our approach significantly outperforms the baselines in the segmentation metrics and achieves the lowest error on the stereo task.

We present a real-time system for 3D head pose estimation and facial landmark localization using a commodity depth sensor. We introduce a novel triangular surface patch (TSP) descriptor, which encodes the shape of the 3D surface of the face within a triangular area. The proposed descriptor is viewpoint invariant, and it is robust to noise and to variations in the data resolution. Using a fast nearest neighbor lookup, TSP descriptors from an input depth map are matched to the most similar ones that were computed from synthetic head models in a training phase. The matched triangular surface patches in the training set are used to compute estimates of the 3D head pose and facial landmark positions in the input depth map. By sampling many TSP descriptors, many votes for pose and landmark positions are generated which together yield robust final estimates. We evaluate our approach on the publicly available Biwi Kinect Head Pose Database to compare it against state-of-the-art methods. Our results show a significant improvement in the accuracy of both pose and landmark location estimates while maintaining real-time speed.

We propose a method to rank and retrieve image sequences from a natural language text query, consisting of multiple sentences or paragraphs. One of the method's key applications is to visualize visitors' text-only reviews on TRIPADVISOR or YELP, by automatically retrieving the most illustrative image sequences. While most previous work has dealt with the relations between a natural language sentence and an image or a video, our work extends to the relations between paragraphs and image sequences. Our approach leverages the vast user-generated resource of blog posts and photo streams on the Web. We use blog posts as text-image parallel training data that co-locate informative text with representative images that are carefully selected by users. We exploit large-scale photo streams to augment the image samples for retrieval. We design a latent structural SVM framework to learn the semantic relevance relations between text and image sequences. We present both quantitative and qualitative results on the newly created DISNEYLAND dataset.

The availability of accurate depth cameras have made real-time human pose estimation possible; however, there are still demands for faster algorithms on low power processors. This paper introduces 1000 frames per second pose estimation method on a single core CPU. A large computation gain is achieved by random walk sub-sampling. Instead of training trees for pixel-wise classification, a regression tree is trained to estimate the probability distribution to the direction toward the particular joint, relative to the current position. At test time, the direction for the random walk is randomly chosen from a set of representative directions. The new position is found by a constant step toward the direction, and the distribution for next direction is found at the new position. The continual random walk through 3D space will eventually produce an expectation of step positions, which we estimate as the joint position. A regression tree is built separately for each joint. The number of random walk steps can be assigned for each joint so that the computation time is consistent regardless of the size of body segmentation. The experiments show that even with large computation gain, the accuracy is higher or comparable to the state-of-the-art pose estimation methods.

The importance of precise homography estimation is often underestimated even though it plays a crucial role in various vision applications such as plane or planarity detection, scene degeneracy tests, camera motion classification, image stitching, and many more. Ignoring the radial distortion component in homography estimation---even for classical perspective cameras---may lead to significant errors or totally wrong estimates. In this paper, we fill the gap among the homography estimation methods by presenting two algorithms for estimating homography between two cameras with different radial distortions. Both algorithms can handle planar scenes as well as scenes where the relative motion between the cameras is a pure rotation. The first algorithm uses the minimal number of five image point correspondences and solves a nonlinear system of polynomial equations using the Groebner basis method. The second algorithm uses a non-minimal number of six image point correspondences and leads to a simple system of two quadratic equations in two unknowns and one system of six linear equations. The proposed algorithms are fast, stable, and can be efficiently used inside a RANSAC loop.

The popularity of low-cost RGB-D scanners is increasing on a daily basis. Nevertheless, existing scanners often cannot capture subtle details in the environment. We present a novel method to enhance the depth map by fusing the intensity and depth information to create more detailed range profiles. The lighting model we use can handle natural scene illumination. It is integrated in a shape from shading like technique to improve the visual fidelity of the reconstructed object. Unlike previous efforts in this domain, the detailed geometry is calculated directly, without the need to explicitly find and integrate surface normals. In addition, the proposed method operates four orders of magnitude faster than the state of the art. Qualitative and quantitative visual and statistical evidence support the improvement in the depth obtained by the suggested method.

We present a minimal, non-iterative solution to the absolute pose problem for images from rolling shutter cameras. Absolute pose problem is a key problem in computer vision and rolling shutter is present in a vast majority of today's digital cameras. We propose several rolling shutter camera models and verify their feasibility for a polynomial solver. A solution based on linearized camera model is chosen and verified in several experiments. We use a linear approximation to the camera orientation, which is meaningful only around the identity rotation. We show that the standard P3P algorithm is able to estimate camera orientation within 6 degrees for camera rotation velocity as high as 30deg/frame. Therefore we can use the standard P3P algorithm to estimate camera orientation and to bring the camera rotation matrix close to the identity. Using this solution, camera position, orientation, translational velocity and angular velocity can be computed using six 2D-to-3D correspondences, with orientation error under half a degree and relative position error under 2%. A significant improvement in terms of the number of inliers in RANSAC is demonstrated.

Feature fusion has been proven effective [31, 32] in image search. Typically, it is assumed that the to-be-fused heterogeneous features work well by themselves for the query. However, in a more realistic situation, one does not know in advance whether a feature is effective or not for a given query. As a result, it is of great importance to identify feature effectiveness in a query-adaptive manner. Towards this goal, this paper proposes a simple yet effective late fusion method at score level. Our motivation is that the sorted score curve exhibits an "L" shape for a good feature, but descends gradually for a bad one (Fig. 1). By approximating score curve's tail with a reference collected on irrelevant data, the effectiveness of a feature can be estimated as negatively related to the area under the normalized score curve. Experiments are conducted on two image search datasets and one person re-identification dataset. We show that our method is robust to parameter changes, and outperforms two popular fusion schemes, especially on the resistance to bad features. On the three datasets, our results are competitive to the state-of-the-arts.

Rapid progress in unconstrained face recognition has resulted in a saturation in recognition accuracy for current benchmark datasets. While important for early progress, a chief limitation in most benchmark datasets is the use of a commodity face detector to select face imagery. The implication of this strategy is restricted variations in face pose and other confounding factors. This paper introduces the IARPA Janus Benchmark A (IJB-A), a publicly available media in the wild dataset containing 500 subjects with manually localized face images. Key features of the IJB-A dataset are: (i) full pose variation, (ii) joint use for face recognition and face detection benchmarking, (iii) a mix of images and videos, (iv) wider geographic variation of subjects, (v) protocols supporting both open-set identification (1:N search) and verification (1:1 comparison), (vi) an optional protocol that allows modeling of gallery subjects, and (vii) ground truth eye and nose locations. The dataset has been developed using 1,501,267 million crowd sourced annotations. Baseline accuracies for both face detection and face recognition from commercial and open source algorithms demonstrate the challenge offered by this new unconstrained benchmark.

Motivated by reasons related to data security and privacy, we propose a method to limit meaningful visual contents of a display from being captured by screenshots. Traditional methods take a system architectural approach to protect against screenshots. We depart from this framework, and instead exploit image processing techniques to distort visual data of a display and present the distorted data to the viewer. Given that a screenshot captures distorted visual contents, it yields limited useful data. We exploit the human visual system to empower viewers to automatically and mentally recover the distorted contents into a meaningful form in real-time. Towards this end, we leverage on findings from psychological studies which show that blending of visual information from recent and current fixations enables human to form meaningful representation of a scene. We model this blending of information by an additive process, and exploit this to design a visual contents distortion algorithm that supports real-time contents recovery by the human visual system. Our experiments and user study demonstrate the feasibility of our method to allow viewers to readily interpret visual contents of a display, while limiting meaningful contents from being captured by screenshots.