Past research in shape analysis and OCR has often emphasized graph matching techniques. We propose to use matching of graph embeddings because this is what is actually of interest. In this way we obtain faster and sim...
详细信息
In this paper we describe the NPen++ system for writer independent on-line handwriting recognition. This recognizer needs no training for a particular writer and can recognize any common writing style (cursive, hand-p...
详细信息
We describe a system that automatically identifies the script used in documents stored electronically in image form. The system can leam to distinguish any number of scripts. It develops a set of representative symbol...
This paper addresses a novel recognition technique for rotated character strings. Rotated character strings appear in character/graph mixed drawings such as cadastral maps, engineering drawings, etc. Recently, there a...
详细信息
A novel intraframe source coding algorithm suitable for the recording of digital High Definition Television signals is presented. A multi-layered, hierarchical description of the source signal is obtained by means of ...
详细信息
A novel intraframe source coding algorithm suitable for the recording of digital High Definition Television signals is presented. A multi-layered, hierarchical description of the source signal is obtained by means of a quad-tree, half-band wavelet transform. This transform decomposes the input signal into a collection of spectrally non-overlapping subbands. Individual subbands are quantised and entropy coded by using a novel predictive arithmetic coding technique. The algorithm is tuned to achieve bit-rate reduction ratios in the range 8:1 - 4:1 which is most useful for recording applications. Results obtained from simulating the coding algorithm, show noticeable improvement over the current state-of-the-art international standard algorithm for still picture encoding both in terms of subjective quality and of measured mean-square error.
The determination of 3-D motion is a very important task particularly in mobile robots control and car driving assistance. Most of the approaches employed consists of two major steps;isolation of the objects of the sc...
详细信息
The determination of 3-D motion is a very important task particularly in mobile robots control and car driving assistance. Most of the approaches employed consists of two major steps;isolation of the objects of the scene, and the estimation of the motion components of the objects. Several researchers have studied the possibility of recovering the 3-D motion without point to point correspondence. In one of this studies, a two-step procedure was suggested and applied to a sequence of stereoscopic images. In this article, an original system to determine axial motion maps without spatio-temporal matching is described. In this approach, objects are not isolated before motion analysis, and the axial motion transformation is the only factor considered.
In this paper, we propose a new scheme for multiresolution recognition of totally unconstrained handwritten numerals using wavelet transform and a simple multilayer cluster neural network. The proposed scheme consists...
详细信息
Fitting surfaces to 3-D data is one of the basic methods for surface description for 3-D vision. Most techniques of surface fitting proposed in the literature are 'least-squares'-based that rarely produce sati...
详细信息
Fitting surfaces to 3-D data is one of the basic methods for surface description for 3-D vision. Most techniques of surface fitting proposed in the literature are 'least-squares'-based that rarely produce satisfactory results if a certain level of noise is present in the data or if the data points are locally sampled from a small area. We propose a new approach that minimizes the mean squared approximate orthogonal distances with linearization using the Newton iteration method. This approach usually yields a good fit and the algorithm is reliable and efficient for real applications. Results are reported for one of the real range images that we have experimented. The results demonstrate that the approximate orthogonal distance performs better than the least squares based methods.
The neural network learning algorithm presented in the paper splits the problem of handwritten digit recognition into easy steps by learning character classes incrementally: At each step, the neurons most relevant to ...
详细信息
A method is dEveloped to extract lines from pages of handwritten text by finding the shortest spanning tree of a graph formed from the set of main strokes. Main strokes of extracted lines are arranged in the same orde...
详细信息
暂无评论