Mohsen Rashwan | Cairo University (original) (raw)

Papers by Mohsen Rashwan

The last two decades witnessed some advances in the development of an Arabic character recognitio... more The last two decades witnessed some advances in the development of an Arabic character recognition (CR) system. Arabic CR faces technical problems not encountered in any other language that make Arabic CR systems achieve relatively low accuracy and retards establishing them as market products. We propose the basic stages towards a system that attacks the problem of recognizing online Arabic cursive handwriting. Rule-based methods are used to perform simultaneous segmentation and recognition of word portions in an unconstrained cursively handwritten document using dynamic programming. The output of these stages is in the form of a ranked list of the possible decisions. A new technique for text line separation is also used.

Page 1. Hierarchical Classification of Bank Checks Using Genetic Algorithms Heba A. Elnemr Mohsen... more Page 1. Hierarchical Classification of Bank Checks Using Genetic Algorithms Heba A. Elnemr Mohsen Rashwan Mohammed S. Elsherif Ahmed Hussien ... 1 2.2. Patron data localization - i Fig. 1-c Fig. ld mean value of each region are found and thresholded to a certain values. ...

$Research paper thumbnail of Matching criteria in fractal image compression$

In video compression, the motion compensation phase passes through block matching step. During th... more In video compression, the motion compensation phase passes through block matching step. During this phase we search in two successive frames about a block in the last one with respect to its original position in the first one and apply mathematical technique to decide if it matches or not. The proposed method is for matching phase, the normal methods are Mean Absolute Difference (MAD), Mean Square Difference (MSD), Pel Difference Classification (PDC) and Integral Projection (IP). We proposed a method based on the subsampling while applying IP then adding the adaptively phase across a preprocessed factor depending on the nature of the frame content.

We proposed a composite adaptive technique for color image compression. We apply a mapping transf... more We proposed a composite adaptive technique for color image compression. We apply a mapping transform from (R, G, B) space to (Yd, Cr, Cb) to represent the image as luminance image. The proposed compression technique is based on discrete cosine transform (DCT) with block size of 8 by 8 pixels. The proposed adaptive DCT technique depends on the maximum luminance value and the difference between this maximum and the minimum one in each block. This will help to select one quantization matrix out of 16 different matrices. The quantization matrices represent the whole range of types of the blocks. As an error correction step we create a 3D matrix that represents the whole image. The X axis is the number of blocks per row, the Y axis is the number of blocks per column and the Z axis is the 256 colors of the image. This will replace the coding of Cr, Cb components and it will solve the problem is case where the luminance value represents more than one color. The technique achieves a high compression ratio in the range of 0.65 bpp for Yd component of the smooth image to 1 bpp for detailed image in addition to overheads in the range from 0.15 to 0.25 bpp for the error correction matrix. The technique gives a higher compression ratio compared ratio compared to the well known technique for standard images, such as Lenna, while maintaining the same quality.

This paper introduces a fast and efficient indexing approach for both 2D and 3D model-based objec... more This paper introduces a fast and efficient indexing approach for both 2D and 3D model-based object recognition in the presence of rotation, translation, and scale variations of objects. The indexing entries are computed after preprocessing the data by Haar wavelet decomposition. The scheme is based on a unified image feature detection approach based on Zernike moments. A set of low level features, e.g. high precision edges, gray level corners, are estimated by a set of orthogonal Zernike moments, calculated locally around every image point. A high dimensional, highly descriptive indexing entries are then calculated based on the correlation of these local features and employed for fast access to the model database to generate hypotheses. A list of the most candidate models is then presented by evaluating the hypotheses. Experimental results are included to demonstrate the effectiveness of the proposed indexing approach.

Modeling phonetic context is one of the key points to get natural sounding in concatenativc speec... more Modeling phonetic context is one of the key points to get natural sounding in concatenativc speech synthesis. In this paper, a new quantitative method to model context is proposed. In the proposed method, the context is measured as the distance between leafs of the top-down likelihood-based decision trees that have been grown during the construction of acoustic inventory. Unlike other context modeling methods, this method allows the unit selection algorithm to borrow unit occurrences from other contexts when their context distances are close. This is done by incorporating the measured distance as an element in the unit selection cost function. The motivation behind this method is that it reduces the required speech modification by using better unit occurrences from near context. This method also makes it easy to use long synthesis units, e.g. syllables or words, in the same unit selection framework

Self learning machines as defined in this paper are those learning by observation under limited s... more Self learning machines as defined in this paper are those learning by observation under limited supervision, and continuously adapt by observing the surrounding environment. The aim is to mimic the behavior of human brain learning from surroundings with limited supervision, and adapting its learning according to input sensory observations. Recently, Deep Belief Nets (DBNs) [1] have made good use of unsupervised learning as pre-training stage, which is equivalent to the observation stage in humans. However, they still need supervised training set to adjust the network parameters, as well as being nonadaptive to real world examples. In this paper, Self Learning Machine (SLM) is proposed based on deep belief networks and deep auto encoders.

signal processing and communications systems. The performance of the FFT component is a key facto... more signal processing and communications systems. The performance of the FFT component is a key factor in evaluating the overall system performance, and it is common to use it as a benchmark for the whole system. Many attempts have been made to enhance the FFT performance, both on algorithm and implementation levels. Software and hardware designs exist to implement this component. In this paper, an optimized hardware implementation of FFT processor on FPGA is presented, where the steps of Radix-2 FFT algorithm are well analyzed and an optimized design is developed as a result, with full exploitation of the hardware platform capabilities to achieve optimum performance. The performance results of the proposed design are demonstrated, and compared to other related works and reference designs.

In this paper, the need for practically solving natural language problems via statistical methods... more In this paper, the need for practically solving natural language problems via statistical methods besides the (incomplete) classical closed form ones is manifested and formalized. Also, the statistical features of natural language are extensively discussed. One simple yet effective work frame of this type of processing that relies on long n-grams probability estimation plus powerful and efficient tree search algorithms is presented in detail. Finally, the authors exemplify the above ideas through a practical Arabic text diacritizer (hence a phonetic transcriptor for Arabic speech applications especially TTS) that has been built according to that work frame

$Research paper thumbnail of Matching criteria in fractal image compression$