Search

Identifying permanent water and temporary water in flood disasters efficiently has mainly relied on change detection method from multi-temporal remote sensing imageries, but estimating the water type in flood disaster events from only post-flood remote sensing imageries still remains challenging. Research progress in recent years has demonstrated the excellent potential of multi-source data fusion and deep learning algorithms in improving flood detection, while this field has only been studied initially due to the lack of large-scale labelled remote sensing images of flood events. Here, we present new deep learning algorithms and a multi-source data fusion driven flood inundation mapping approach by leveraging a large-scale publicly available Sen1Flood11 dataset consisting of roughly 4831 labelled Sentinel-1 SAR and Sentinel-2 optical imagery gathered from flood events worldwide in recent years. Specifically, we proposed an automatic segmentation method for surface water, permanent water, and temporary water identification, and all tasks share the same convolutional neural network architecture. We utilize focal loss to deal with the class (water/non-water) imbalance problem. Thorough ablation experiments and analysis confirmed the effectiveness of various proposed designs. In comparison experiments, the method proposed in this paper is superior to other classical models. Our model achieves a mean Intersection over Union (mIoU) of 52.99%, Intersection over Union (IoU) of 52.30%, and Overall Accuracy (OA) of 92.81% on the Sen1Flood11 test set. On the Sen1Flood11 Bolivia test set, our model also achieves very high mIoU (47.88%), IoU (76.74%), and OA (95.59%) and shows good generalization ability.

This paper presents an approach for image cartoonization. By observing the cartoon painting behavior and consulting artists, we propose to separately identify three white-box representations from images: the surface representation that contains smooth surface of cartoon images, the structure representation that refers to the sparse color-blocks and flatten global content in the celluloid style workflow, and the texture representation that reflects high-frequency texture, contours and details in cartoon images. A Generative Adversarial Network (GAN) framework is used to learn the extracted representations and to cartoonize images. The learning objectives of our method are separately based on each extracted representations, making our framework controllable and adjustable. This enables our approach to meet artists' requirements in different styles and diverse use cases. Qualitative comparisons and quantitative analyses, as well as user studies, have been conducted to validate the effectiveness of this approach, and our method outperforms previous methods in all comparisons. Finally, the ablation study demonstrates the influence of each component in our framework.

Polygonized isosurfaces of anatomical structures commonly suffer from severe artifacts (e.g.,?noise and staircases), due to inhomogeneous binary volumes. Most state-of-the-art techniques can reduce these artifacts but inevitably ruining anatomical structures' morphology. Given an initial polygonization of an isosurface, we first eliminate these apparent staircases based on a context-aware Laplace filter, and then solve the morphology-preserving problem of anatomical structures as an optimization of the local spatial quadrics (LSQ) of fitted Bezier surfaces during mesh evolution. This results in a conceptually simple approach that provides a unified framework for not only handling artifacts, but also for enabling the morphology preservation of anatomical structures. We design an effective mesh smoothing framework that focuses on medical data.The LSQ gives an isosurface mesh a compact approximation to the underlying surface.We improve the staircase detection strategy by a novel cascaded operation.

While modern optical and laser 3D scanners can generate high accuracy mesh models, to largely avoid their introducing noise which prohibits practical applications still results in high cost. Thus, optimizing noisy meshes while preserving their geometric details is necessary for production, which still remains as challenging work. In this paper we propose a novel and efficient two-stage feature-preserving mesh denoising framework which can remove noise while preserving fine features of a surface mesh. We improve the capability of feature preservation of our vertex updating scheme by employing an extension of the quadric error metrics (QEM), which can track and minimize updating errors and hence well preserve the overall shape as well as detailed features of a mesh. We further leverage vertex normals to guide the vertex updating process, as the normal field of a mesh reflects the geometry of the underlying surface. In addition, to obtain a more accurate normal field to guide vertex updating, we develop an improved normal filter by integrating advantages of existing filters. Compared with traditional gradient descent based schemes, our method performs better on challenging regions with rich geometric features. Moreover, a local entropy metric is proposed to measure stability of a mesh and the effectiveness of vertex updating algorithms. Qualitative and quantitative experiments demonstrate that our approach can effectively remove noise from noisy meshes while preserving or recovering geometrical features of original objects.

In this chapter, we examine the global chromatic and achromatic aspects of a painting. Their organisation is particular to each artist, with work by that artist showing consistency in its contrast organisation. This organisation forms a significant part of their ‘style’. Contrast can be broken down into its chromatic aspects (hue and saturation) and its achromatic aspect (lightness). We propose a method by which this style may be transferred from a painting to a photograph. The principle behind our method is to transfer the inherent structures of the contrast properties. This differentiates our method from a simple colour transfer, using which the hues of an image may be shifted, and in doing so damages the semantic properties of the photograph.

The lightness and saturation of an image may be subject to adjustment for the purpose of aesthetic improvement. Normally, this is in the form of a contrast adjustment. Such an adjustment would work upon the maximum, minimum and mean values of the lightness and/or saturation. The third value in the colour triumvirate is hue, but existing image improvement approaches rarely manipulate this value for the reason of its importance to the semantic reading of an image. Looking to painters for inspiration, we propose a method by which a region of contiguous hue in an image may be subject to one of two contrast adjustments: hue spread and hue compression. These are broadly analogous to adjustments of the relative form of Itten’s contrast of temperature. We develop a tool, which employs superpixels, using which a user may segment regions of similar hue, at varying scales. This allows the user to quickly select semantically related regions. These local regions also provide the basic region of operation for three proposed hue operations: (1) hue spread, (2) hue compression and (3) hue shift. We show that when combined with superpixel segmentation, these operations are capable of increasing, decreasing and changing regions of local hue contrast in a manner that is aesthetically agreeable. We demonstrate our framework on a variety of input images and discuss its evaluation from the feedback of several expert users.

In the previous chapter, colour contrast was examined as phenomena within the art domain. In this chapter, we review computational models that might be employed to define these phenomena. In the engineering domain, there is not a standard measurement model of lightness, saturation or hue contrast. All existing models would probably be better regarded as application tools rather than as descriptions of basic visual function. Colour contrast is broadly understood as being one of two things: a relationship between two things (i.e. simple contrast) or the variance of a property within a single thing (i.e. complex contrast).

Quantifying the style of visual art has long been a challenging yet largely overlooked problem. Of key importance in the style of a painting is the manner in which the artist has employed hue. We propose a means to quantify global hue contrast as two values: hue variety and hue antagonism. These were derived from Johannes Itten’s work on colour contrast. Using our approach, the hue contrast of a single image can be visually expressed as a single point in a plot. For this reason, the approach is particularly suitable for the comparative analysis of groups of artworks, such as those by two or more artists or an artist at different times in their career. We describe two case studies of such an analysis, and the results agree with observations. We also include a user study that supports the validity of our proposed approach.

Paper pop-ups are interesting three-dimensional books that fascinate people of all ages. The design and construction of these pop-up books however are done manually and require a lot of time and effort. This has led to computer-assisted or automated tools for designing paper pop-ups. This paper proposes an approach for automatically converting a 3D model into a multi-style paper pop-up. Previous automated approaches have only focused on single-style pop-ups, where each is made of a single type of pop-up mechanisms. In our work, we combine multiple styles in a pop-up, which is more representative of actual artist's creations. Our method abstracts a 3D model using suitable primitive shapes that both facilitate the formation of the considered pop-up mechanisms and closely approximate the input model. Each shape is then abstracted using a set of 2D patches that combine to form a valid pop-up. We define geometric conditions that ensure the validity of the combined pop-up structures. In addition, our method also employs an image-based approach for producing the patches to preserve the textures, finer details and important contours of the input model. Finally, our system produces a printable design layout and decides an assembly order for the construction instructions. The feasibility of our results is verified by constructing the actual paper pop-ups from the designs generated by our system.

Ultrasonography is an important tool and has been widely used in clinical applications, however, the physicians and surgeons still often suffers great difficulties in diagnosis and treatment due to the high speckle noise of ultrasound images. Existing speckle reduction methods usually over-smooth low contrast features since they are sensitive to contrast variations in the images. In our paper, we propose a novel learning-based scheme to effectively remove the speckle noise while preserving the image features. This is achieved by incorporating a denoising algorithm — non-local simultaneous sparse coding (NLSSC) which is modified by adding a pre-clustering process guided by feature asymmetry (FA). The proposed pre-clustering procedures can help better train the dictionaries in NLSSC and make them more feature-aware. Qualitative and quantitative experiments were carried out to evaluate the performance of our methods against some state-of-the-art on both synthetic and clinical ultrasound images. Experimental results demonstrate that our approach has effective performance and can produce good speckle reduced images.

One of the most common ways to adjust the aesthetic quality of a natural scene image is by editing its color. However, while lightness and saturation manipulation is well understood in computational terms, hue manipulation is poorly understood and non-trivial to affect. As a result, modifications made to hue with existing photo-editing tools often impart an unnatural look to the image. In this paper, we discuss a framework for hue manipulation and its contribution to the aesthetics of an image inspired by the use of hue in the visual arts. Our framework is designed to work with a segmentation approach based on superpixels that can segment the image into regions of similar color at varying scales. This allows the user the ability to quickly select semantically similar regions. These local regions also provide the basic region of operation for three proposed hue operations: 1 hue spread; 2 hue compression; and 3 hue shift. We show that when combined with superpixel segmentation these operations are capable of increasing, decreasing and changing regions of local hue contrast in a manner that produces natural or artistic looking images. We demonstrate our framework on a variety of input images and discuss its evaluation from the feedback of several expert users.

Most mesh denoising techniques utilize only either the facet normal field or the vertex normal field of a mesh surface. The two normal fields, though contain some redundant geometry information of the same model, can provide additional information that the other field lacks. Thus, considering only one normal field is likely to overlook some geometric features. In this paper, we take advantage of the piecewise consistent property of the two normal fields and propose an effective framework in which they are filtered and integrated using a novel method to guide the denoising process. Our key observation is that, decomposing the inconsistent field at challenging regions into multiple piecewise consistent fields makes the two fields complementary to each other and produces better results. Our approach consists of three steps: vertex classification , bi-normal filtering , and vertex position update . The classification step allows us to filter the two fields on a piecewise smooth surface rather than a surface that is smooth everywhere. Based on the piecewise consistence of the two normal fields, we filtered them using a piecewise smooth region clustering strategy. To benefit from the bi-normal filtering, we design a quadratic optimization algorithm for vertex position update. Experimental results on synthetic and real data show that our algorithm achieves higher quality results than current approaches on surfaces with multifarious geometric features and irregular surface sampling.

Accurate segmentation of brain tumor from MR image is crucial for the diagnosis and treatment of brain cancer. We propose a novel automated brain tumor segmentation method based on a probabilistic model combining sparse coding and Markov random field (MRF). We formulate the brain tumor segmentation task as a pixel-wise labeling problem with regard to three classes: tumor, edema and healthy issue. For each class, dictionary learning is performed independently on multi-modality gray scale patches. Sparse representation is then extracted based on a joint dictionary which is constructed by combing the three independent dictionaries. Finally, we build the probabilistic model aiming to estimate maximum a posterior (MAP) probability by introducing the sparse representation into likelihood probability and prior probability using the Markov random field (MRF) assumption. Compared with traditional methods, which employed hand-crafted low level features to construct the probabilistic model, our model can better represent the characteristics of a pixel and its relation with neighbors based on the sparse coefficients obtained from the learned dictionary. We validated our method on the MICAAI 2012 BRATS challenge brain MRI dataset and achieved comparable or better results compared with state-of the-art methods.

Color quantization is a technique for decreasing the color depth of an image. Less number of color means less space is needed for storage. Such kind of algorithms ([Heckbert 1982], [Verevka and Buchanan 1995]) have invoked a big interest in the computer science domain especially for researchers in the network field, since color reduction could benefit hardware and network resources. However, few of the existing color reduction algorithms consider preservation of image details while reducing color depth.

This paper presents an interactive web-based anatomy navigation system based on the high-resolution Chinese Visible Human (CVH) dataset. Compared with previous anatomy learning software, there are three new features in our navigation system. First, we directly exploit the capabilities of graphics hardware to achieve real-time computation of large medical dataset on the web. In addition, various visualization effects are supplied to enhance the visual perception of human model. Second, to facilitate user interaction, we design a set of user-friendly interface by incorporating the Microsoft Kinect into the system, and the users can navigate the Visible Human with their hand gestures. Third, in order to eliminate the unreliable bottleneck: network transmission, we employ a progressive strategy to transmit the data between the server and the client. Experimental results validate the advantages of the proposed navigation system for learning human anatomy, indicating its great potential in clinical applications.