International Journal of Image and Graphics

A new concept, referred to as rayset, is discussed in this paper. It is a parametric function consisting of two mapping relations. The first one maps from a parameter space to the ray space, while the second one maps from the parameter space to the attribute space. A taxonomy is proposed based on the rayset concept whereby scene representations and scene reconstruction techniques used in image-based rendering are individually classified. Existing image-based rendering techniques are surveyed under the proposed classification. The review shows that different image-based scene representations, such as multiple-center-of-projection image and concentric mosaics, can be cast as different kinds of raysets. Different scene reconstruction approaches can be regarded as attempts to render different raysets. The concepts of rayset warping and rayset editing are also formulated. Under the rayset taxonomy, both techniques try to alter one of the mapping relations defined by a rayset without changing the other one.

Reconstruction of freeform surfaces from scattered coordinate data is a difficult problem encountered in many surface fitting and geometric modeling applications. Conventional tessellation and parametric surface fitting techniques are limited because they require prior knowledge about the connectivity between the sampled points. The method of surface reconstruction described in this paper exploits the learning capability of a self-organizing feature map (SOFM) to adaptively fit a deformable sphere to the unorganized 3D coordinate data. The learning algorithm automatically establishes the connectivity between the measured points by iteratively changing the topological relationships within the map. By incorporating additional constraints during the learning process it is possible to have the deformable map follow the shape of objects with surface holes and cavities. Several examples of freeform surfaces with varying levels of complexity are discussed in order to demonstrate the performance of the algorithm.

In the field of content-based image retrieval, there exist a gap between low-level descriptions of image content and the semantic needs of users to query image databases. This paper demonstrates an approach to image retrieval founded on classifying image regions hierarchically based on their semantics (e.g. sky, snow, rocks, etc.) that resemble peoples' perception rather than on low-level features (e.g. color, texture, shape, etc.). Particularly, we consider outdoor images and automatically classify their regions based on their semantics using a support vector machines (SVMs). The SVMs learns the semantics of specified classes from specific low-level feature of the test image regions. Image regions are, first, segmented using a hill-climbing approach. Then, those regions are classified by the SVMs. Such semantic classification allows the implementation of intuitive query interface. As we show in our experiments, the high precision of semantic classification justifies the feasibility of our approach.

Traditional image enhancement algorithms do not account for the subjective evaluation of human operators. Every observer has a different opinion of an ideally enhanced image. Automated Techniques for obtaining a subjectively ideal image enhancement are desirable, but currently do not exist. In this paper, we demonstrate that Reinforcement Learning is a potential method for solving this problem. We have developed an agent that uses the Q-learning algorithm. The agent modifies the contrast of an image with a simple linear point transformation based on the histogram of the image and feedback it receives from human observers. The results of several testing sessions have indicated that the agent performs well within a limited number of iterations.

This paper presents a semi-automatic registration algorithm for partially overlapped aerial images which has been successfully tested with images of the Mississippi Delta area. In this algorithm, each individual aerial image is registered by employing a pattern search method. This pattern search method considers an affine transformation without shear as a 5-parameter vector and searches toward the gradient direction that results in higher similarity values between the reference and the sensed images. The registration of the first aerial image requires some initial manual work. After the registration of the first image, based on the overlap of the neighboring images and the existing transformation parameters for the first image, the search starting point of the second image can be automatically obtained for the registration of the second image. This process can be repeated for the remaining aerial images in a sequential order. Experimental results with 12 sample images demonstrate the success of this algorithm.

Most parameter-based online signature verification methods achieve correspondence between the points of two signatures by minimizing the accumulation of their local feature distances. The matching based on minimizing the local feature distances alone is not adequate since the point contains not only local features but the distribution of the remaining points relative to it. One useful way to get correspondences between points on two shapes and measure the similarity of the two shapes is to use the shape context, since this descriptor can be used to describe the distributive relationship between a reference point and the remaining points on a shape. In this paper, we introduce a shape context descriptor for describing an online signature point which contains both 2D spatial information and a time stamp. A common algorithm, dynamic time warp (DTW), is used for the elastic matching between two signatures. When combining shape contexts and local features, we achieve better results than when using only the local features. We evaluate the proposed method on a signature database from the First International Signature Verification Competition (SVC2004). Experimental results demonstrate that the shape context is a good feature and has available complementarity for describing the signature point. The best result by combining the shape contexts and the local features yields an Equal Error Rate (EER) of 6.77% for five references.

In this paper, a new 2D shape Multiscale Triangle-Area Representation (MTAR) method is proposed. This representation utilizes a simple geometric principle, that is, the area of the triangles formed by the shape boundary points. The wavelet transform is used for smoothing and decomposing the shape boundaries into multiscale levels. At each scale level, a TAR image and the corresponding Maxima-Minima lines are obtained. The resulting MTAR is more robust to noise, less complex, and more selective than similar methods such as the curvature scale-space (CSS). Furthermore, the MTAR is invariant to the general affine transformations. The proposed MTAR is tested and compared to the CSS method using MPEG-7 CE-shape-1 part B and Columbia Object Image Library (COIL-20) datasets. The results show that the proposed MTAR outperforms the CSS method for the conducted tests.

This paper presents a new methodology for the deformation of soft objects by drawing an analogy between Poisson equation and elastic deformation. The potential energy stored in an elastic body as a result of a deformation caused by an external force is propagated among mass points by Poisson equation. An improved Poisson model is developed for propagating the energy generated by the external force in a natural manner. A cellular neural network (CNN) model is established to solve the Poisson model for the real-time requirement of soft object deformation. A haptic virtual reality system has been established for deformation simulation with force feedback. This proposed methodology not only deals with local and large-range deformations, but also accommodates isotropic, anisotropic and inhomogeneous materials by simply modifying constitutive coefficients, as well as predicts the mechanical behaviors of living tissues.

In this paper, we present a novel method for the optimal texture sampling and texture synthesis using vector quantization clustering. The synthesizing process selects an initial texture block from the input sample texture and then iteratively searches all the texture patches from the sampled texture space for a patch whose boundary best matches the current boundary of the synthesized texture. To control the number of the searching patches while preserving the seamless visual quality, we present two effective patch sampling techniques, namely uniformly distributed patch sampling and patch quantization clustering. The sampled texture patches will then be further optimized using principle component analysis. In this scheme, the complexity of the patch search in each step can be significantly reduced and thus the seamless texture synthesis can be performed in real-time. Experimental results show that the proposed schemes are of excellent efficiency for the visually seamless texture synthesis of a wide variety of textures ranging from regular to stochastic can be achieved.

A new feature based image mosaic method is presented in this paper. Our main contribution is an improved RANSAC algorithm based on the modified median flow filter to find the inliers and the homography between the images. This algorithm overcomes the limit of local feature detection and thus improves the stability of homography calculation. Except for the RANSAC algorithm, PCA-SIFT is used to detect features because it is more compact, more robust and faster than the standard SIFT. The priority search is also used to match the feature points in a faster way than BBF search. Experiments show that our method is robust for image mosaic.