L. Wang and T. Pavlidis, #Direct gray-scale extraction of features for character recognition," IEEE Trans. on Pattern Analysis and Machine Intelligence 15, pp. 1053#1067, October 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....are with Graduate School of Engineering, Tohoku University, Sendai shi, 9808579 Japan. M. Inoue is with Hitachi Software Engineering Co. Ltd. Yokohama shi, 231 0015 Japan. Fig. 1. Examples of decorated characters. a) A logotype. b) Various fonts of character A. intensity surface [12] [13], 14] 15] are extracted from multi scale images. Ridges are used for global structure extraction, and ravines are used for interpolation. Experimental results show clear character structures are extracted from very complex decorated characters. Moreover, the effectiveness of the algorithm is ....

L. Wang and T. Pavilidis, "Direct Gray-Scale Extraction of Features for Character Recognition," IEEE Trans. Pattern Analysis and Machine Intelligence, vol.15, no.10, pp.1053--1067, Oct. 1993.

....to its concavity region. This approach is currently investigated for the automatic recognition of car plates. Note that our approach does not require to threshold the characters beforehand. Hence, it fits well the direct grey scale extraction of features for character recognition recommended in [22]. 5. Concluding remarks The definition of the convex hull of a set in terms of an intersection of morphological closings with a series of half planes allows us to extend the convex hull transform to grey scale images. It is also at the basis of a very efficient implementation. The extension of ....

L. Wang and Y. Pavlidis. Direct gray-scale extraction of features for character recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(10):1053--1066, October 1993.

....has been proposed to implement this definition. The extension of our work to 3 D grey scale data is possible using an approach similar to that developed in [11] for line segments. Application of convex hulls to extracting features directly from grey scale images for character recognition purposes [33] has been introduced in [27] Acknowledgments I would like to thank the anonymous referees for their stimulating comments and Nigel McFarlane for helpful discussions. A. Convex Superset Approximation by One Pixel Let us consider two pixels at coordinates (0, 0) and (x,y) We assume that x and ....

L. Wang and Y. Pavlidis. Direct gray-scale extraction of features for character recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(10):1053-1066, October 1993.

....how gray level processing can be used to improve the system performance. Topographic analysis of gray scale images was originally proposed by Haralick et al. 4] Topographic analysis has been shown to be better than binary analysis for symbol recognition in an automatic mail sorting application [5], 6] Topographic analysis extracts the information in the gray scale image by computing topographic labels for each pixel, and uses this to segment touching symbols or merge fragmented symbols. Binary analysis is based on a locally adaptive thresholding of the gray scale image. In [5] 6] a ....

....[5] 6] Topographic analysis extracts the information in the gray scale image by computing topographic labels for each pixel, and uses this to segment touching symbols or merge fragmented symbols. Binary analysis is based on a locally adaptive thresholding of the gray scale image. In [5], 6] a fairly simple locally adaptive binarization method [7] was used. Recent evaluation studies [8] 9] suggested that Niblack s method [10] augmented with the postprocessing step of Yanowitz and Bruckstein [11] is a very good locally adaptive binarization method for hydrographic maps. It is ....

[Article contains additional citation context not shown here]

L. Wang and T. Pavlidis, "Direct gray-scale extraction of features for character recognition", IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 15, no. 10, pp. 1053--1067, Oct. 1993.

....and ravines are detected by locally approximating the image function by its second or third order Taylor polynomial. The coefficients of this polynomial are usually determined by using the facet model, i.e. by a least squares fit of the polynomial to the image data over a window of a certain size [2, 15, 6] or by using derivatives of Gaussian masks [9, 4] The approaches using the facet model have the problem that only lines of a certain width can be extracted [14] while the Gaussian masks can be tuned for a certain line width by selecting an appropriate oe. It is also possible to select the ....

L. Wang and T. Pavlidis. Direct gray-scale extraction of features for character recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(10):1053--1067, Oct. 1993. 7

....and ravines are detected by locally approximating the image function by its second or third order Taylor polynomial. The coefficients of this polynomial are usually determined by using the facet model, i.e. by a least squares fit of the polynomial to the image data over a window of a certain size [20, 21, 22, 23, 24, 25]. The direction of the line is determined from the Hessian matrix of the Taylor polynomial. Line points are then found by selecting pixels that have a high second directional derivative perpendicular to the line direction. The advantage of this approach is that lines can be detected with sub pixel ....

....when the sub pixel location of each line point is taken into account it can be seen that there is always a single response to a given line since all line point locations line up prefectly. Therefore, linking will be considerably easier than in approaches that yield multiple responses, e.g. [27, 21, 22], and no thinning operation is needed [34] 3 Linking Line Points into Lines After individual line pixels have been extracted, they need to be linked into lines. It is necessary to do this right after the extraction of the line points because the later stages of determining line 10 width and ....

Li Wang and Theo Pavlidis. Direct gray-scale extraction of features for character recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(10):1053-- 1067, October 1993. 28

....line segments to the strokes to obtain the width and direction of the stroke at that point. Their reason for using grayscale features rather than converting the picture to black and white (for example by thresholding) is to avoid irreversible, uninformed decisions [3] 4 Wang and Pavlidis [15] use a gradient maxima technique to extract a clearer black and white stroke description by assuming the grayscale word is composed of a topological map of pixel intensities at each pixel location. Their results show a clearer stroke representation than plain dynamic thresholding. Abuhaiba, Holt ....

Li Wang and Theo Pavlidis. Direct gray-scale extraction of features for character recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(10):1053--1067, October 1993.

....confidence in the input data. But when the area is ambiguous, we defer the decision to later processing stages. We pass on to later stages, not only a graph description, but also various ambiguities of the input image which will be handled later by the primary classifier, on a knowledge based way [82]. ii) Knowledge based shape interpretation using graph matching Graph Matching seems to be a flexible and powerful tool for shape interpretation, since it can absorb shape distortions and provide us with strong evidence about the identity of the input character. At a first stage, feature ....

L. Wang and T. Pavlidis. Direct gray scale extraction of features for character recognition. IEEE Trans. on Pattern Anal. and Machine Intel., 15:1053--1067, Oct. 1993.

....sign of a Gaussian is derived for each point on the surface of a range image. Image segmentation is then achieved through the identification of primitive critical points (peaks, pits, ridges, etc. Haralick et al. 1983) defined a similar set of topographic features for use in 2D image analysis, Wang and Pavlidis (1993), and later Lee and Kim (1995) extended this work to extract features for character recognition. Gauch and Pizer (1993) also identify ridge and valley bottoms in 2D images, where a ridge is defined as a point where the intensity falls off sharply in two directions and a valley bottom is a point ....

Wang, L., & Pavlidis, T. (1993). Direct gray scale extraction of features for character recognition. IEEE Trans. Patt. Anal. Mach. Intell., PAMI-15 (10), 1053--1067.

....methods are described, respectively. In section 4, experiment results are shown, and concluding remarks are followed in section 5. 2. TOPOGRAPHIC FEATURE CLASSIFICATION One approach of topographic feature classification is to adopt mathematical framework about 3 dimensional surface description[1 3]. For derivative operation on the discrete image, it usually uses several convolution operations, which is time consuming. The other approach is to classify pixels according to intensity configuration of neighborhood and appropriate heuristic decision rules [4 6] The gray level of a center pixel ....

....convolution [3] 3. STROKE EXTRACTION In real digital images, collection of skeletal pixels does not satisfy one pixelwide connectivity. Therefore, a sophisticated stroke extraction method is needed. In Wang and Pavlidis work, a set of graph structured strokes is extracted from skeletal pixels [1]. An edge is a chain of lines obtained using analysis of the adjacent ridge pixel region. Each of adjacent peaks and saddles is represented as a node on a graph. Toriwaki and Fukumura applied a thinning algorithm on the collection of ridge and peak pixels to obtain line representation [6] Since ....

L. Wang and T. Pavlidis, "Direct Gray-Scale Extraction of Features for Character Recognition," IEEE Trans. PAMI, vol. 15, no. 10, pp. 1053-1067,

....assume that the characters appear in the same text string and have known orientation. In hydrographic maps (Fig. 1) for example, some characters touch or overlap lines, or touch characters from another text line. Trier et al. 52] have developed a method based on gray scale topographic analysis [53, 54], which integrates binarization and segmentation. This method gives a better performance, since information gained in the topographic analysis step is used in segmenting the binary image. The segmentation step also handles rotated characters and touching characters from different text strings. The ....

....A character graph can be derived from the skeleton by approximating it with a number of straight line segments and junction points. Arcs may be used for curved parts of the skeleton. Wang and Pavlidis have recently proposed a method for obtaining character graphs directly from the gray level image [53, 68]. They view the gray level image as a 3D surface, with the gray levels mapped along the z coordinate, using z = 0 for white (background) and, for example, z = 255 for black. By using topographic analysis, ridge lines and saddle points are identified, which are then used to obtain character graphs ....

L. Wang and T. Pavlidis, "Direct gray-scale extraction of features for character recognition, " IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 15, pp. 1053--1067, Oct. 1993.

....without performing any preprocessing (binarisation, skeletonisation) to extract features. This scheme has the advantage that it does not introduce distortion or spurious artifacts into the word image. There has been a recent trend towards the use of grey scale images, for example Wang and Pavlidis [51] who used the grey scale image to generate the skeleton, but there has been very little on the extraction of features directly from the grey scale image itself. Two new methods for determining the global slope of a word image were presented. Of the two, the Fourier domain approach was prefered as ....

L. Wang, T. Pavlidis, "Direct Gray-Scale Extraction of Features for Character Recognition", IEEE Trans. on Pattern Analysis and Machine Intelligence, vol. 15, no. 10, pp. 1053-1067, 1993.

.... (compared with arbitrary images and graphics) civilized problem of text recognition (OCR) 14] Early work with template matching was followed with more robust shape or feature based techniques [15, 16, 17] More recently the importance of preserving gray scale information has been recognized [18, 19]. Lexical and contextual knowledge [20] has been shown by many to further improve the accuracy of recognition algorithms. One extensive study [21] demonstrated that the best commercial software achieved only 98.6 accuracy, which means that substantial human post editing is still required and that ....

L.Wang and T. Pavlidis, "Direct gray-scale extraction of features for character recognition," PAMI, 15(10):1053--1067, October 1993.

....how gray level processing can be used to improve the system performance. Topographic analysis of gray scale images was originally proposed by Haralick et al. 4] Topographic analysis has been shown to be better than binary analysis for symbol recognition in an automatic mail sorting application [5], 6] Topographic analysis extracts the information in the gray scale image by computing topographic labels for each pixel, and uses this to segment touching symbols or merge fragmented symbols. Binary analysis is based on a locally adaptive thresholding of the gray scale image. In [5] 6] a ....

....application [5] 6] Topographic analysis extracts the information in the gray scale image by computing topographic labels for each pixel, and uses this to segment touching symbols or merge fragmented symbols. Binary analysis is based on a locally adaptive thresholding of the gray scale image. In [5], 6] a fairly simple locally adaptive binarization method [7] was used. Recent evaluation studies [8] 9] suggested that Niblack s method [10] augmented with the postprocessing step of Yanowitz and Bruckstein [11] is a very good locally adaptive binarization method for hydrographic maps. It is ....

[Article contains additional citation context not shown here]

Li Wang and Theo Pavlidis, "Direct gray-scale extraction of features for character recognition", IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 15, no. 10, pp. 1053--1067, Oct. 1993.

....at avoiding these segmentation problems by working directly on the grey level image and perform recognition word by word instead of character by character. Little work has been done in the area of text and character recognition from grey level images, but two examples can be found in [1] and [2]. The latter also gives an overview of previous work in this field. The method we have used in this study is very simple and straight forward. Word recognition has been payed more attention, and in [3] 4] 5] some studies based on the use of hidden Markov models (HMMs) can be found. We have ....

L. Wang & T. Pavlidis: "Direct Gray-Scale Extraction of Features for Character Recognition" IEEE Trans. Pattern Machine Intell., Vol. 15, No. 10, 1993

No context found.

L. Wang and T. Pavlidis, #Direct gray-scale extraction of features for character recognition," IEEE Trans. on Pattern Analysis and Machine Intelligence 15, pp. 1053#1067, October 1993.

No context found.

L. Wang and T. Pavlidis, "Direct Gray-Scale Extraction of Features for Character Recognition", IEEE Transactions on Pattern Analysis and Machine Intelligence , Vol. 10, pp. 1053-1067, October 1993.

No context found.

L. Wang and T. Pavlidis. Direct gray scale extraction of features for character recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15:1053--1067, October 1993.

No context found.

L. Wang and T. Pavlidis, "Direct gray-scale extraction of features for character recognition," IEEE Trans. Pattern Anal. Machine Intell., vol. 15, pp. 1053--1067, Oct. 1993.

No context found.

L. Wang and T. Pavlidis, "Direct Gray-Scale Extraction of Features for Character Recognition," IEEE Trans. on Pattern Anal. and Mach. Intell., vol. 15, no. 10, pp. 1053--1067, October 1993.

No context found.

Li Wang, Theo Pavlidis, "Direct Gray-Scale Extraction of Features for Character Recognition",IEEE Trans. on Pattern Analysis and Machine Intelligence, vol.15, no.10, pp.10531067, 1993.

No context found.

L. Wang and T. Pavilidis, "Direct Gray-Scale Extraction of Features for Character Recognition," IEEE Trans. Pattern Analysis and Machine Intelligence, vol.15, no.10, pp.1053--1067, Oct. 1993.

No context found.

L. Wang and T. Pavlidis, "Direct gray-scale extraction of features for character recognition", IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 15, no. 10, pp. 1053--1067, Oct. 1993.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC