Results

**11 - 16**of**16**### Hybrid Fractal/DCT Coding Of Video

, 1995

"... In this paper we introduce two techniques that can be used together to improve video coding performance. Firstly, block matching motion estimation with spatial transformation (BMST) is used to increase motion compression efficiency, and then motion compensated prediction (MCP) errors are encoded usi ..."

Abstract
- Add to MetaCart

In this paper we introduce two techniques that can be used together to improve video coding performance. Firstly, block matching motion estimation with spatial transformation (BMST) is used to increase motion compression efficiency, and then motion compensated prediction (MCP) errors are encoded using a hybrid Fractal/DCT scheme. MCP error blocks are classified according to their energy content and they should be coded with a low bit rate. Many of them can be DCT coded, but for those which require a large bit rate to achieve a certain quality, Fractal coding can be efficiently used considering the bit rate constraint. This hybrid scheme speeds up the decoding process, reduces the bit rate, and due the efficiency of motion compensation only a few MCP error images are coded. INTRODUCTION Video compression techniques use essentially two types of redundancy reduction techniques: interframe coding, to exploit the temporal redundancy between consecutive frames, and intraframe coding, to ...

### Connectionist Pyramid Powered Perceptual Organization: Visual Grouping with Hierarchical Structures of Neural Networks

, 1995

"... This paper describes a new approach for organizing image data. Perceptual organization in biological vision is the process of pre-attentively grouping and structuring perceptual information into shapes and forms. In computer vision, processes like these are required in order to transform pixels into ..."

Abstract
- Add to MetaCart

This paper describes a new approach for organizing image data. Perceptual organization in biological vision is the process of pre-attentively grouping and structuring perceptual information into shapes and forms. In computer vision, processes like these are required in order to transform pixels into structural forms that can be used in higher-level interpretation. Pyramidal structures have been used for perceptual organization in computer vision. The tapering of the structure provides a basis for the "bringing together" of spatially separate image elements. Artificial neural networks are a means of a system learning by example rather than being pre-coded with the necessary rules. (A problem with performing perceptual organization with computer systems is that biological perceptual organization is not that well understood.) This study focussed on the grouping of line segments into longer lines. A multi-layer feed-forward neural network is successfully trained with backpropagation to per...

### 708 T. MARKAS and J. REIF ENCODER DECODER

"... arithmetic coding [6], or Lempel-Ziv type methods [7,8] to attain a lossless technique that exceeds the performance of stand-alone lossless methods. 1.1 Vector quantization Vector quantization is a lossy block-coding technique that is used extensively to compress data at low bit rates (high compress ..."

Abstract
- Add to MetaCart

arithmetic coding [6], or Lempel-Ziv type methods [7,8] to attain a lossless technique that exceeds the performance of stand-alone lossless methods. 1.1 Vector quantization Vector quantization is a lossy block-coding technique that is used extensively to compress data at low bit rates (high compression ratios). Vector quantizers compress image data by replacing an image block, represented by a discrete vector, with the index of the bestmatch entry found in a reconstruction codebook based on a given distortion measure [9]. A reconstruction codebook consisting of vectors of size N divides the entire search space into a fixed number of N-dimensional regions. The encoding process is responsible for identifying the corresponding region of each input block and for replacing the block with the index of the codebook vector that represents that region (Fig. 2). This is accomplished by measuring the distortion between the input block and each codebook vector, and by identifying the vector with the minimum distortion. The decoding process is a simple look-up operation where each index of the compressed data string is used to rebuild the original image. Both the encoder and the decoder maintain the same vocabulary that has been constructed off-line using training algorithms on a set of data that is similar to the one that will be compressed. Some of the training algorithms that can be used to construct the codebook include the widely used k-means algorithm [lo], and neural network type algorithms. A straightforward implementation of the vector quantization algorithm is to perform a linear search over all vocabulary entries. This full search method is a computationally expensive task since linear time is required to compare each input vector against all codebook entries using a given distortion measure. A more efficient approach that reduces the computational requirements from linear to logarithmic time is the tree-structured vector quan-

### Segmentation Coding Using Edge Detection and Region Merging

"... An algorithm that integrates edge detection and region merging is presented. The algorithm can be employed as a pre-processing operation for model based image coding schemes. The pixels of the same intensities in an image are firstly clustered into small regions, resulting an oversegmented image whe ..."

Abstract
- Add to MetaCart

An algorithm that integrates edge detection and region merging is presented. The algorithm can be employed as a pre-processing operation for model based image coding schemes. The pixels of the same intensities in an image are firstly clustered into small regions, resulting an oversegmented image where the boundaries are mainly edges. The further region merging operations are conducted in a way that the parameters are set to produce segmented image suitable for transform coding. Computer simulations he show that this algorithm yields suitable segmentation for effective transform coding. 1.

### COMPUTER VISION, GRAPHICS, AND IMAGE PROCESSING x,207-216 (1984) Octree Representations of Moving Objects*

, 1982

"... An algorithm is described that updates an object’s octree representation as the object is linearly translated through space. This is accomplished by performing simple arithmetic on the path representations of the nodes to be translated. Among others, one advantage of the algorithm is in devising col ..."

Abstract
- Add to MetaCart

An algorithm is described that updates an object’s octree representation as the object is linearly translated through space. This is accomplished by performing simple arithmetic on the path representations of the nodes to be translated. Among others, one advantage of the algorithm is in devising collision-free and efficient trajectories of moving objects in robotics. 1.

### COMPUTER VISION, GRAPHICS, AND IMAGE PROCESSING?vd, 200-214 (1983) On Approaches to Polygonal Decomposition for Hierarchical Image Representation NARFNDRA

, 1982

"... Approaches to polygonal decomposition for hierarchical image representation are described. For planar decomposition, quad trees using square and triangular neighborhoods are found to be the only feasible methods, having the same computational complexity. For grid images the choice of the appropriate ..."

Abstract
- Add to MetaCart

Approaches to polygonal decomposition for hierarchical image representation are described. For planar decomposition, quad trees using square and triangular neighborhoods are found to be the only feasible methods, having the same computational complexity. For grid images the choice of the appropriate tree type is determined by the grid topology. Triangular and square quad trees are appropriate for the triangular and square grids, whereas trees of order 7 are necessary for the hexagonal grid. 1.