International Journal of Image and Graphics

There have been many applications in the area of handwritten character recognition. Over the last decade much research has gone into developing algorithms to accurately convert captured images of handwriting to text. At the same time, research into neuro fuzzy classification models has proven to solve many complex problems. In this paper, Adaptive Neuro Fuzzy Inference System (ANFIS) and Evolving Fuzzy Neural Network (EFuNN) was investigated and studied in detail on how these two models can be used to perform handwritten digits classification. Results of the experiments show great potential of the EFuNN over the ANFIS for practical implementation of the handwritten digit recognition.

Video compression techniques have been applied routinely in order to conserve storage space and minimize bandwidth utilization in various video applications. To reduce inter-pixel redundancies inside and between video frames, video codecs (coder/decoder or compression/decompression) are mainly characterized by applying motion estimation and motion compensation (MEMC) in combination with discrete cosine transform (DCT). However, blocking artifacts are obvious from the block-based motion estimation and transformation, especially at low bit rates. Due to the intrinsic advantages of multiresolution and scalability of discrete wavelet transform (DWT), video compression techniques have been infused with exciting prospects by the beneficial integration of motion estimation and wavelet transformation. The contributions of this paper are in (1) proposing a design architecture of the wavelet-domain encoder and decoder; (2) implementing the multiresolution motion estimation and compensation method; and in (3) realizing the dynamic entropy encoding and decoding, so as to achieve more efficient video compression than the conventional spatial domain methods for low bit rate applications without the annoying blocking artifacts.

We propose studying the variations of volume of implicit objects during an animation according to several points of view: choice of the function of density, variations of parameters such as the iso-value and the radius of influence for a given function, variations of the parameters inherent in a particular function. Modification of parameters of the function of density must be carried out with care. There are no rules concerning these variations. To avoid the non-monotonous variations, it is necessary to choose a function of density beforehand and study the intervals of variation of its parameters. A new discretization makes it possible to locate these variations for a later use in a process of control of these variations.

A new image segmentation method is developed that combines the advantage of the normalized cuts (Ncut) algorithm to solve the perceptual grouping problem by means of graph partitioning, and the ability of wavelet transform to capture image features by decomposing signal both in time and frequency. We derive image features from orientation histograms defined on the detail subbands of the discrete wavelet transform. The segmentation is implemented by partitioning a graph representing an image at the coarsest transform level, while the weights of the graph are calculated from all the scales. Due to the reduced dimensionality of the dataset, the speed of Ncut is increased. Even though segmentation is carried out at a coarsest level of transform, the results are accurate for images of different structural contents, including textures.

In this paper, two novel texture descriptors and two novel edge descriptors are proposed, which are low-dimension, effective, and are obtained by a relative simple approach. The two texture descriptors are the directional difference unit and the gradient unit histogram, which are rotation invariant. The two edge descriptors are the quantized max-min difference histogram and the quantized fuzzy entropy histogram, the former is more suitable for describing the images with relatively regular texture characteristic, the latter is more suitable for describing the images with relatively regular structure characteristic or no regular characteristic. Combing color descriptor, texture descriptor and edge descriptor to form hybrid visual feature index to retrieve natural color images, this method is insensitive to image rotation and translation. Experimental results show that the method achieves better performance than other recent relevant methods.

Algorithms for rendering numerous virtual characters often achieve interactive frame rate by reducing the computation load. The major drawback of this approach is that the rendering quality is degraded. This project aims to develop a fast rendering scheme that can be used on realistic virtual human characters. We have made two contributions. Firstly, we have created a high quality human model and designed three basic deformation methods. The deformation of each body part is composed of the basic deformation methods. Our aim is to generate visually similar deformed and seamless character efficiently. Rendering of numerous characters at high frame rate can be maintained by using impostor. It is well known that impostor has the visibility problem. Secondly, to correct the visibility problem, we propose to incorporate the depth information to each character by using stencil buffer and bounding box. Our rendering scheme compares favorably, in terms of the frame rate and the rendering quality, with multi-layered impostor.

Many problems solved in computer graphics, computer vision, visualization etc. require fast and robust computation of an area of a triangle or volume of a tetrahedron. These very often used algorithms are well known and robust if vertices coordinates of triangles or tetrahedrons are given in Euclidean coordinates. The homogeneous coordinates are often used for the representation of geometric transformations. They enable us to represent translation, rotation, scaling and projection operations in a unique way and handle them properly. Today's graphics hardware based on GPU offers very high computational power using pixel shaders and fragment shaders not only for graphical elements processing, but also for general computation using GPU as well.

This paper presents simple methods for the area of a triangle and the volume of a tetrahedron computation if vertices are given in homogeneous coordinates without the need to use the division operation for vertices coordinates transformation from the homogeneous coordinates to the Euclidean coordinates. Area or volume computation is transferred to the cross product computation that is fast, simple, and robust and can be supported in hardware or implemented on GPU that uses vector operations with homogeneous coordinates natively. The presented formula can be used directly for Euclidean representation just setting w equal to 1.

Relief texture mapping is an image-based rendering technique which can successfully support the representation of 3D surface details and view motion parallax. It has the potential to significantly increase visual realism of rendered geometry while keeping system load constant. In this paper, FPGA (Field Programmable Gate Array) chip technology is applied to this three-dimensional image warping method. A relief texture mapping system has been implemented on a reprogrammable and reconfigurable FPGA board. The algorithm is optimized for the specific architecture and the framework is customized for circuit resources, which can be flexibly changed for other structures. In our design, we take advantage of inherent parallelism of the algorithm by concatenating multiple warping engines and well organizing data in memory space. Experimental results show high image quality with improved rendering speed.

Harmonic interpolation has been shown to produce smooth interpolating curves when applied to points in 2D or 3D space. The techniques used in harmonic interpolation can be applied equally well to image data. In this paper, harmonic interpolation is used to interpolate the colors in an image. We identify certain properties of harmonic interpolation which allows us to transform the image into another image more suitable for conventional compression techniques to be applied. The properties identified also allow for further compression of the image in addition to conventional compression techniques. We apply the algorithm to a variety of images and present the results.

In this paper, we propose a new algorithm for removing Gaussian noise from natural images while preserving important details for further processing. Similar patches appear frequently in an image. Our algorithm is based on this high redundancy in natural images. Our initial step is to perform an image segmentation by a constrained quad-tree-split-and-merge technique. Then for each patch, using an extended SSD, we identify all similar patches in the image and recalculate the patch by summing up a weighted copy of each similar patch. Since the mean of white noise is zero, the final image will be noise free and important details are preserved. Experimental comparisons of Wiener Filter, Wavelet Threshold Methods and our method will be carried out by using different types of images. The result shows that our method is more efficient in both noise removing and detail preserving.