Nikolaos Papamarkos

Multithresholding of mixed-type documents

Mixed type documents include text, drawings and graphics regions. It is obvious that a technique that can reduce the number of the gray-levels in accordance to the type of each document region could be important for many document applications, such as storage, transmission and recognition. To solve this problem this paper proposes a new method that is called the document multithresholding technique. The method is based on a Page Layout Analysis (PLA) technique and on a neural network multilevel threshold selection approach. In the final document the different block types are stored with the appropriate and limited number of gray-level values. In text and line-drawing blocks, only one threshold is determined whereas in the graphics blocks the optimal number of thresholds is first determined. The performance of the method was extensively tested on a variety of documents.

Document image binarization using local features and Gaussian mixture modeling

In this paper, we address the document image binarization problem with a three-stage procedure. First, possible stains and general document background information are removed from the image through a background removal stage. The remaining misclassified background and character pixels are then separated using a Local Co-occurrence Mapping, local contrast and a two-state Gaussian Mixture Model. Finally, some isolated misclassified components are removed by a morphology operator. The proposed scheme offers robust and fast performance, especially for both handwritten and printed documents, which compares favorably with other binarization methods. Background removal technique based on adaptive median filtering and thresholdingA Local Co-occurrence Map with local contrast can distinguish between document text and document stains and background.Low complexity approach with fast and accurate binarization results

An intelligent hardware structure for impulse noise suppression

In this paper an intelligent hardware module suitable for the computation of an adaptive median filter (AMF) is presented. The proposed digital hardware structure is pipelined and parallel processing is used to minimize computational time. It is capable of processing gray-scale images of 8-bit resolution with 3/spl times/3 or 5/spl times/5-pixel image neighborhoods as options for the computation of the filter output. However, the system can be easily expanded to accommodate windows of larger sizes. The function of the proposed circuitry is to detect the existence of impulse noise in an image neighborhood and apply the median filter operator only when necessary. Moreover, the noise detection procedure can be customized so that a range of pixel values is considered as impulse noise. In this way, the integrity of edge and detail information of the image under process is preserved and blurring is avoided. The proposed digital structure was implemented in FPGA and it can be used in industrial imaging applications, where fast processing is of the utmost importance. As an example, the time required to perform filtering of a grayscale image of 260/spl times/244 pixels is approximately 7.6 msec. The typical system clock frequency is 65 MHz.

Multi-spectral document image binarization using image fusion and background subtraction techniques

In this paper, the authors exploit a multispectral image representation to perform more accurate document image binarisation compared to previous color representations. In the first stage, image fusion is employed to create a “document” and a “background” image. In the second stage, the FastICA algorithm is used to perform background subtraction. In the third stage, a spatial kernel K-harmonic means classifier binarizes the FastICA output. The proposed system outperforms previous efforts on document image binarization.

A gray-scale Inverse Hough Transform algorithm

This paper proposes a gray-scale Inverse Hough Transform. The proposed algorithm can be applied to any gray-scale image and allows the edge extraction according to filtering conditions. The algorithm exploits all the gray-scale information and avoids conversion of the gray-scale image to a binary one. The edge pixels are exactly determined according to their position and gray-level as they appear in the original gray-scale image.

Text identification in color documents

In complex color documents, text, drawings and graphics are appeared with millions of different colors. In many cases, text regions are overlaid onto drawings or graphics. In this paper, a new method is proposed to automatically detect and extract text in mixed type color documents. The proposed method is based on a combination of an adaptive color reduction (ACR) technique and a page layout analysis (PLA) approach. The ACR technique is used to obtain the optimal number of colors. Then, image is split to separable binary images, each one corresponding to every principal color. The PLA technique is applied independently to each one of the color plains and identifies the text regions. A merging procedure is applied in the final stage to merge the text regions derived from the color plains and to produce the final document.

Local Co-occurrence and Contrast Mapping for Document Image Binarization

Document Image Binarization refers to the task of transforming a scanned image of a handwritten or printed document into a bi-level representation containing only characters and background. Here, we address the historic document image binarization problem using a three-stage methodology. Firstly, we remove possible stains and noise from the document image by estimating the document background image. The remaining background and character pixels are separated using a Local Co-occurrence Mapping, local contrast and a two-state Gaussian Mixture Model. In the last stage, possible isolated misclassified blobs are removed by a morphology operator. The proposed scheme offers robust and fast performance, especially for handwritten documents.

Applying conformal geometry for creating a 3D model spatial-consistent texture map

The aim of this research is to achieve spatial consistency of the UV map. We present an approach to produce a fully spatially consistent UV mapping based on the planar parameterisation of the mesh. We apply our method on a 3D digital replica of an ancient Greek Lekythos vessel. We parameterise the mesh of a 3D model onto a unit square 2D plane using computational conformal geometry techniques. The proposed method is genus independent, due to an iterative 3D mesh cutting procedure. Having now the texture of a 3D model depicted on a spatially continuous two dimensional structure enables us to efficiently apply a vast range of image processing based techniques and algorithms.

Text extraction using component analysis and neuro-fuzzy classification on complex backgrounds

This paper proposes a new technique for text extraction on complex color documents and cover books. The novelty of the proposed technique is that contrary to many existing techniques, it has been designed to deal successfully with documents having complex background, character size variations and different fonts. The number of colors of each document image is reduced automatically into a relative small number (usually below ten colors) and each document is divided into binary images. Then, connected component analysis is performed and homogenous groups of connected components (CCs) are created. A set of features is extracted for each group of CCs. Finally each group is classified into text or non-text classes using a neuro-fuzzy classifier. The proposed technique can be summarized into four consequent stages. In the first stage, a pre-processing algorithm filters noisy CCs. Afterwards, CC grouping is performed. Then, a set of nine local and global features is extracted for each group and finally a classification procedure detects documents text regions. Experimental results prove the efficiency of the proposed technique, which can be further extended to deal with even more complex text extraction problems.

Grayscale image reconstruction from projections with linear noise response

This paper presents concisely two reconstruction algorithms that provide exact image reconstruction from its projections and then analyzes the performance of these algorithms when noisy projection data are present. In both the reconstruction methods the projection samples are stored in a 2-D accumulator array where each column corresponds to the projection data at a certain view angle. However, the second method uses a significantly smaller number of projection samples. The response of the reconstruction methods is examined when Gaussian noise is applied in the projection data stored in the accumulator array. Several cases are examined regarding the original image size, the level of the input noise and the rounding of the grayscale values in the reconstructed image. The experimental results show that the two reconstruction methods have the same noise response and that in both cases the reconstructed image is linearly related to the input noise.