Suman K. Mitra

Technique for fractal image compression using genetic algorithm

A new method for fractal image compression is proposed using genetic algorithm (GA) with an elitist model. The self transformation property of images is assumed and exploited in the fractal image compression technique. The technique described utilizes the GA, which greatly decreases the search space for finding the self similarities in the given image. This article presents theory, implementation, and an analytical study of the proposed method along with a simple classification scheme. A comparison with other fractal-based image compression methods is also reported.

A Morphology Based Approach for Car License Plate Extraction

Locating the car license plate in an image or video frame of a car is an important step in car license plate recognition/identification applications. This problem poses many challenges like location of license plate from images taken in poor illumination and bad weather condition; plates that are partly obscured by dirt and images that have low contrast. This paper presents a new morphology based method for license plate extraction from car images. The algorithm uses morphological operations on the preprocessed, edge images of the vehicles. Characteristic features such as license plate width and height, character height and spacing are considered for defining structural elements for morphological operations. Connected component analysis is used to select the band containing license plate from the candidate segmented. The experimental results with a reasonably large set of car images are very encouraging.

Iterative function system and genetic algorithm-based EEG compression

A method for EEG compression is proposed, using Iterative Function System (IFS) and Genetic Algorithms (GAs) with elitist model, keeping the quality sufficiently good for clinical purposes. Compression using IFS is usually called fractal compression. The self transformability property of the EEG signals is assumed and is exploited in the fractal compression technique. To ascertain the self transformability of the EEG signal, some isometric transformations have been applied. The technique described here utilizes Genetic Algorithm that decreases the search space for finding the self similarities in the given signal. This article presents theory and implementation of the proposed method. The fidelity of the reconstructed signal obtained by the present compression algorithm has been assessed both qualitatively and quantitatively. The compression ratios, for the EEG signals in various states, are found to be comparable to the other available techniques for EEG compression. In our method at least 85% data reduction has been achieved.

Rough set based image denoising for brain MR images

In this paper, we propose a novel approach to explore self-similarity of an image for patch based image processing application. The motivation of this work is to search for a similar set of pixels from a given image for each pixel or patch present in the image. So far, the search for similarity exploration in the image is a time consuming task and restricted to a local search space in many of the previous works. The proposed method explores the image space globally for each given patch using Rough Set Theory (RST) in a principled way. The similarity in the image space is explored according to a predefined set of attribute(s) of the image. The selection strategy using RST has been applied for an image denoising task to enhance the capability of the underlying method. We have demonstrated the suitability of RST for a similar patch selection applying it on two state-of-the-art methods and hence proposed a new algorithm in comparison to the state-of-the-art methods that is efficient in terms of computational complexity. The applicability of denoising methods has been shown on the medical image domain and evaluated quantitatively using various statistical measures. The performance of proposed method was found to be comparable and satisfactory. HighlightsThe soft computing, in the precise Rough Set Theory (RST) based Medical Image Denoising problem, is addressed in the work.RST based framework is proposed to find similar patches from the single image for denoising purpose.The proposal is to release the local search constraint for similar patches from state-of-the-art methods.The performance of all the proposals has been compared with state-of-the-art methods and found to be satisfactory.

Text Extraction from Document Images Using Edge Information

Detection of text from documents in which text is embedded in complex colored document images is a very challenging problem. There are a lot of potential uses of text extraction in image searching, archiving documents etc. In this paper, we propose a simple edge based feature to perform this task. It aims at detecting textual regions from the document and separating it from the graphics portion. The algorithm is based on the sharp edges of the characters which are missing in images. We find these edges and use them to classify text from images. This edge information can also be used for other image interpretation tasks.

Human Action Recognition Using DFT

Action is any meaningful movement of the human and it is used to convey information or to interact naturally without any mechanical devices. Human action recognition is motivated by some of the applications such as video retrieval, Human robot interaction, to interact with deaf and dumb people etc. In any Action Recognition System, some pre-processing steps are done for removing the noise caused because of illumination effects, blurring, false contour etc. Background subtraction is done to remove the static or slowly varying background. In this paper, multiple background subtraction algorithms are tested and then one of them is selected for the further process of action recognition. Background subtraction is also known as foreground/background segmentation or foreground extraction. The next step is the feature extraction which deals with the extraction of the important feature (like corner points, optical flow, shape, motion vectors etc.) from the image frame. The proposed novel action recognition algorithm uses discrete Fourier transform (DFT) of the small image block.

A Bayesian network based sequential inference for diagnosis of diseases from retinal images

We propose a system that learns from the STARE (STructured Analysis of REtina) database and exploits the experience of ophthalmologists to assist in decision-making regarding the presence or absence of retinal diseases. The developed system automatically detects diseases given a description (a set of manifestations) of a retinal image. The manifestations in the retinal image are usually fed sequentially into the system where the manifestation dependences and order must be learned by the system. We apply naive Bayes classifier which is a simple case of Bayesian network to learn the conditional probabilities and to establish an approximate lookup table for sequential manifestation input. The system interacts with the ophthalmologist in determining the sequence of manifestations for inferring the correct disease. The overall performance of the system is found to be satisfactory and useful by ophthalmologists.

A technique for image magnification using partitioned iterative function system

Abstract A new technique for image magnification using the theory of fractals is proposed. The technique is designed assuming self-transformability property of images. In particular, the magnification task is performed using the fractal code of the image instead of the original one resulting in a reduction in memory requirement. To generate the fractal codes, Genetic Algorithm with elitist model is used which greatly decreases the search for finding self similarities in the given image. The article presents both theory and implementation of the proposed method. A simple distortion measure scheme and a similarity criterion based on just noticeable difference have also been proposed to judge the image quality of the magnified image. Comparison with one of the most popular magnification techniques, the nearest-neighbor technique, is made.

On some variants of locality preserving projection

High dimensional data is hard to interpret and work with in its raw form; hence dimensionality reduction is applied beforehand to discover underlying low dimensional manifold. Locality Preserving Projection (LPP) was introduced using the concept that neighboring data points in the high dimensional space should remain neighbors in the low dimensional space as well. In a typical pattern recognition problem, true neighbors are defined as the patterns belonging to same class. Ambiguities in regions having data points from different classes close by, less reducibility capacity and data dependent parameters are some of the issues with conventional LPP. In this article, some of the variants of LPP have been introduced that try to resolve these problems. A weighing function that tunes the parameters depending on data and takes care of the other issues is used in Extended version of LPP (ELPP). Better class discrimination is obtained using the concept of intra and inter-class distance in a supervised variant (ESLPP-MD). To capture the non-linearity of the data, Kernel based variants are used, that first map the data to feature space. Data representation, clustering, face and facial expression recognition performances are reported on a large set of databases.

Motion estimation based color transfer and its application to color video compression

In this paper a novel scheme for color video compression using color transfer technique is proposed. Towards this, a new color transfer mechanism for video using motion estimation is presented. Encoder and decoder architectures for the proposed compression scheme are also presented. In this scheme, compression is achieved by firstly discarding chrominance information for all but selected reference frames and then using motion prediction and discrete cosine transform (DCT) based quantization. At decompression stage, the luminance-only frames are colored using chrominance information from the reference frames applying the proposed color transfer technique. To integrate color transfer mechanism with hybrid compression scheme a new color transfer protocol is defined. Both compression scheme and color transfer work in YCbCr color space.