K. Etemad, D. Doermann, and R. Chellappa, "Multiscale Document Page Segmentation Using Soft Decision Integration," IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 19, no. 1, pp. 92-96, Jan. 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....local features. If a small mask is chosen, it is difficult to detect large scale textures such as large fonts. On the contrary, if a large mask is chosen, the computational cost will increase dramatically. To avoid this problem, the multiscale analysis has been applied in recent researches [8] [9]. In these methods, local characteristics can be extracted at each resolution with only two filters. However, the process of getting multiscale texture information is quite time consuming and, in some cases, regions of different types but with similar texture can be confused or merged. In ....

K. Etemad, D. Doermann, and R. Chellappa, "Multiscale Document Page Segmentation Using Soft Decision Integration," IEEE Trans. Pattern Analysis and Machine Intdligence, vol. 19, pp. 92-96, 1997.

....of our document image compression scheme effectively. Work done by Sennhauser and Ohnesorge [94] and Sauvola and Pietikainen [93] proves that block based compression gives better performance on document images that contain images. It is also useful to consider algorithms such as Etemad et al. [25], Anatoncopoulos et al. 7] Jain et al. 45] Pavlidis et al. 80] and others [28, 79, 108] to provide image segmentation into regions of different types so that different compression schemes may operate on the regions. We assume that there exist enough components in the text regions that can be ....

K. Etemad, D. Doermann, and R. Chellappa. Multiscale document page segmentation using soft decision integration. IEEE Transactions on Pattern Analysis and Machine Intelligence, 19:92--96, 1997.

....degrees of membership of each pixel in the labels are computed. A two layer neural network combines the membership values using the fl model [101] as the aggregation operator, which allows varying degrees of compensation between the union and intersection. Local Decision Integration Etemad et al. [20, 21] present a document page segmentation algorithm which classifies regions into text, image, and graphics. The algorithm uses feature vectors based on a multi scale wavelet packet representation for local classification. Segmentation is done by propagating soft local decisions made on small windows ....

K. Etemad, D. S. Doermann, and R. Chellappa, "Multiscale document page segmentation using soft decision integration," Technical Report CAR-TR-761/CSTR -3444, Center for Automation Research, University of Maryland, March 1995.

No context found.

K. Etemad, D. Doermann, and R. Chellappa, "Multiscale Document Page Segmentation Using Soft Decision Integration," IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 19, no. 1, pp. 92-96, Jan. 1997.

....clustering techniques used in [9] 10] An artificial neural network is a natural choice as a classifier because of its ability to learn. Theoretically, a three layer neural network can approximate any nonlinear function after training. The success of neural networks in related problems [11], 12] 13] provides us with further motivation to rely on a neural network as a classifier to identify text regions. To facilitate the detection of various text sizes, we use a pyramid of images generated from the original image by halving the resolution at each level. The extracted text regions ....

K. Etemad, D. S. Doermann, and R. Chellappa, "Multiscale document page segmentation using soft decision integration," IEEE Trans. PAMI, vol. 19, pp. 92--96, 1997.

....type (list, table, or drawing) or Logical Features which are class dependent, such as names, titles and abstracts, are examples. 3.1. 4 Structural Features A more advanced system evolves when the DA (Document Analysis) page decomposition algorithms also give initial classifications of the regions [4, 20]. This can be regarded as a pre classification of page structure. Most algorithms classify extracted 10 xxxxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx xxxxxxxx xxxxxx xxxx xxxx xxxx xxxx Document Structure Analysis ....

K. Etemad, D. Doermann, and R. Chellappa. Multiscale document page segmentation using soft decision integration. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1996. To appear.

....of documents which have similar column structure and similar layout of graphics regions. They also use the similarity measures to identify title like pages by image example. 18 4. 2 The Use of Texture Texture features have been used extensively for page segmentation and zone classification [34, 68, 33, 15, 24] and have recently been used for image categorization and retrieval. Soffer [54] attempted to extend the concept of n grams (used extensively for noisy text data) to images by extracting NxM grams of image intensity patterns. The technique essentially codes any given image as a set of small ....

K. Etemad, D. Doermann, and R. Chellappa. Multiscale document page segmentation using soft decision integration. IEEE Transactions on Pattern Analysis and Machine Intelligence, 19(1):92--96, 1997.

No context found.

K. Etemad, D. Doermann, and R. Chellappa. Multiscale document page segmentation using soft decision integration. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1996. To appear.

....1. The architecture for extracting text from video frames. have different texture properties than the surrounding areas. This texture reflects common frequency and orientation information making wavelets a reasonable candidate for representation. The success of neural networks in related problems [2] provides us the motivation to rely on a hybrid wavelet neural network segmenter approach to classify text regions. To allow the detection of various text sizes, we use a pyramid of images generated from original image by halving the resolution at each level. The extracted text regions are ....

K. Etemad, D. Doermann, and R. Chellappa. Multiscale document page segmentation using soft decision integration. IEEE Transactions on Pattern Analysis and Machine Intelligence, 19(1):92--96, 1997.

....text ( Pi) and nontext ( in the subspace after LDA projection. Text and nontext overlap. choice as a classifier because of its ability to learn. Theoretically, a three layer neural network can approximate any nonlinear function after training. The success of neural networks in related problems [8, 11, 13, 39, 48] provides us with further motivation to rely on a neural network as a classifier to identify text regions. Our methodology uses a small window (typically 16 Theta 16) to scan the image and classify each window as text or non text using the neural network (Figure 2) We will address the following ....

K. Etemad, D. S. Doermann, and R. Chellappa. Multiscale document page segmentation using soft decision integration. IEEE Trans. PAMI, 19:92--96, 1997.

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