Biologically Rationalized Computing Techniques For Image Processing Applications

Attempts to classify high-resolution satellite data with conventional classifier show limited success since the traditional-per-pixel classifiers examine only the spectral variance ignoring the spatial distribution of the pixels corresponding to the land use/land cover classes. The work is carried out in two stages on panchromatic sharpened IRS P-6 LISS-IV (2.5 m) multispectral (MS) imagery of the year 2014 of Mangalore coastal zone along the west coast of Karnataka state of India. In the first stage, in order to overcome the limitations experienced in the parametric and nonparametric classifications, the swarm intelligence optimisation technique based on Artificial Bee Colony (ABC) algorithm has been studied for twelve land cover classes that are mapped. In the second stage, to bring out a greater separability between the spectrally overlapping classes, a texture-based image classification approach has been introduced and a methodology is developed to determine the optimal window size, interpixel distance and the best combinations of texture bands in multispectral data. The five texture measures, viz. entropy (ENT), angular second moment (ASM), contrast (CON), MEAN and homogeneity (Hmg) derived from the grey-level co-occurrence matrix (GLCM), are investigated in the study. The major observations and contributions of this work are as follows: in the first stage, the image classifier employing the ABC algorithm exhibits higher classification accuracy when compared with maximum likelihood classifier. In the second stage, the results show that combining textural features and spectral bands in classification approach has proven very useful in delineating the spectrally overlapping classes, particularly at higher class hierarchy level.

Owing to the development of multimedia technology, it is mandatory to perform image compression, while transferring an image from one end to another. The proposed method directly highlights the salient region in WHT domain, which results in the saliency map with lesser computation. The WHT-based saliency map is directly used to guide the image compression. Initially, the important and less important regions are identified using WHT-based visual saliency model. It significantly reduces the entropy and also reserves perceptual fidelity. The main aim of the proposed method is to produce the high-quality compressed images with lesser computational effort and thereby achieving high compression ratio. Due to the simplicity and high speed of WHT, the proposed visual saliency-based image compression method is producing reliable results, in terms of peak signal-to-noise ratio (PSNR), compression ratio, and structural similarity (SSIM), compared to the state-of-the-art methods.

This paper presents a prototype called HOLOTECH that implements a model prediction using a limited description of the environment to support blind people. The goal is to perform fast detection and identification of obstacles to provide information about collision riskiness and location. The prediction is not probabilistic but statistic, in order to improve the inferences results. The model works fine using low-precision images drifted from real-time camera of a regular Android cell phone supported with ultrasonic sensors. The main focus of this work is how to pre-process images on the fly in order to be able to train and to tune the plastic learning module, improving the object’s trajectory prediction.

We addressed an algorithm intuitively modeling multi-view information for pattern recognition application, specifically for electromyography (EMG) classification. The objective of the framework is to extract the low-dimensional embeddings (LDEs) inherent in multiple views that comprehensively represent the class information. We have shown that the algorithm is capable of providing robust solution to multitask learning relying on multi-view information. On two sets of EMG data, the learned LDEs comprehensively represent the multi-view information they were trained to represent, with consistency in performance across multiple sets of partitioned data sets. The significant aftermaths of the adopted learning strategy affirm the practical usability of the algorithm in healthcare applications for making correct diagnosis. Further, implementation of the algorithm for spectrogram image-based recognition is also of note.

Stroke is the cerebrovascular issue influencing blood supply to the mind that predominantly influences individuals over 65 years old. This article proposes an automatic technique to perceive and orchestrate the sorts of strokes starting with 2D cerebrum CT images. The methodology is divided into four steps. In the introductory step, preprocessing may be performed on the image to expel unwanted disturbance by applying median filtering. In second step, different texture-based features are extricated utilizing wavelet packet transform (WPT) for classification. In the following step, Linear Discriminant Analysis (LDA) is utilized to diminish the dimensionality of the features. Finally, the diminished group of feature is connected to the supervised learning techniques for classification of normal and infected region. The goal of the proposed work is to build up a framework that accurately extracts the stroke region from CT images that helps doctors in their diagnosis decisions. The performance of the proposed scheme has fundamentally enhanced the stroke classification precision contrasted with other neural system-based classifier.

Over the past few decades due to the advancement of technology, biometrics has evolved into a key factor of security for societal needs. Biometrics started as a basic ID verification system and has evolved into a major factor of authentication by taking several biological parameters as references. There are certain issues that researchers have yet been facing against hack or overriding. This chapter covers a comparative study of different biometrics and its methods used for safe authentication that improves security management for complicated scenarios.

Hyperspectral sensors collect images in hundreds of narrow, continuous spectral channels, whereas multispectral sensors collect images in relatively broader wavelength bands. The spatial resolution of a hyperspectral image is, however, lower than that of a multispectral image. This study has integrated the high spectral and high spatial information of hyperspectral (Hyperion) and multispectral (LISS-IV) imagery of the Henry Island of Sundarbans, India. An integrated image has been successfully generated that has 5 m spatial resolution of LISS-IV image and 10 nm spectral resolution of Hyperion image. The prime objective of this study is to obtain an image of enhanced spectral and spatial resolution that would facilitate accurate interpretation and mapping of mangrove types, sea and creek water, pisciculture water, agricultural soil and saline soil of the study area. The methodology adopted considers band remapping of different spectral regions covered by multispectral and hyperspectral images. This study has applied algorithms to restore spatial information of hyperspectral data by integrating hyperspectral bands with those multispectral bands that fall within its range. After comparison of the spectral profile of fused and original hyperspectral image, similarity is found in the spectral curves across the electronic spectrum indicating that the spectral signature of the fused image is maintained. High values of crosscorrelation between the images show that both the spectral and the spatial information are well preserved. The quality evaluation of the fused image is based on quantitative criteria which include spatial and spectral properties that are defined in the image. This study assesses the quality of fused image through utilization of various statistical indicators. Spectral quality of integrated image is assessed using parameters such as spectral discrepancy, correlation coefficient, values of root mean square error (RMSE), Spectral Angle Mapper (SAM), standard deviation and signal-to-noise ratio (SNR) are used for spectral quality assessment of the fused image. Spatial quality assessment has been done using bias and Edge Detection techniques. After analysis of application of these indicators on the fused and the original hyperspectral image, it is observed that the spectral discrepancy shows low discrepancy values in the visible and near-infrared (NIR) part of the electromagnetic spectrum. The correlation coefficient has given good results in the NIR region as compared to visible region. The results obtained from RMSE show that its values in the visible region are lower than that in the NIR region. SAM performs well in the NIR region as compared to the visible region. The standard deviation of fused and hyperspectral images is very similar and their differences are either close to zero or negative. The values of SNR calculated between fused and hyperspectral images vary randomly with the spectral wavelength. The random variability of the signal over the image (i.e. noise) is very less, thus giving a good SNR value. The result obtained from bias shows a good-quality fused and hyperspectral image with the ideal bias values varying with the spectral wavelength of the fused image. The results thus obtained after application of the indices show that the spectral and spatial property of fused image is close to its ideal value. Neural network has been used to assign classes to pixels of the new integrated image.

Breast cancer has the second highest mortality rate in women next to lung cancer. As per clinical statistics, 1 in every 8 women is diagnosed with breast cancer in their lifetime. However, periodic clinical checkups and self-tests help in early detection and thereby significantly increase the chances of survival. Invasive detection techniques cause rupture of the tumor, accelerating the spread of cancer to adjoining areas. Hence, there arises the need for a more robust, fast, accurate, and efficient noninvasive cancer detection system. In this work, an automated system is proposed for achieving error-free detection of breast cancer using mammogram. In this system, the deep learning techniques such as convolutional neural network, sparse autoencoder, and stacked sparse autoencoder are used. The performance of these techniques is analyzed and compared with the existing methods. From the analysis, it is observed that the stacked sparse autoencoder performs better compared to other methods.

The facial paralysis is a disease that occurs due to facial nerve damage. It is the 7th cranial nerve which resides just behind the ear of human body. Facial nerve travels through a narrow canal, bonny canal called “fallopian canal” in skull, and exits from the skull via the stylomastoid foramen and then passes through parotid gland and subdivided to supply the facial muscles. Facial paralysis is a neuromuscular disorder which involves muscle of expression of the face and taste buds. The common form of facial paralysis is Bell’s palsy that is somehow having same symptoms, but it affects exact half portion of face, i.e., either left side of face which includes eyes, cheeks, lips, and chin or right side of face. But in facial paralysis, one quadrant of face gets affected, i.e., either eyes or lips. In examination of facial paralysis, the face is divided into four quadrants and one or more quadrant may be affected as per the degree of facial paralysis. That is why maximum research defined Bell’s palsy as facial palsy. But there is difference between the Bell’s palsy and facial palsy. It is believed that the various grading system helps to evaluate the degree of facial paralysis in the patients. Incidence of facial paralysis of idiopathic disorder is 25 per 100,000 annually or about 1 in 60 persons in a lifetime. The rate of growth of patients increased with the age of human being. In the USA, it affects about 40,000 people every year. As rate of growth of facial paralysis patients increasing per year so, Computer Science Engineers are participating in this for betterment in recovery and accuracy using grading system and 3D image processing which is not yet used in facial paralysis area. Paper is used to present all common techniques for detecting the facial paralysis and calculating the degree of facial paralysis by using grading system. Paper represents the comparison of all the common grading systems and different problem under one umbrella. Main objective of this paper is to present a review on the progress of research efforts made in various grading systems and then analyzing the scaling methods. The commonly used House–Brackmann grading system (HBGS) is discussed which can help to find the degree of facial paralysis. Higher the grade, more is the paralysis. The development of new method for detection and grading facial nerve function will be very challenging, but it will help the facial paralyzed patients up to some extent.

Generally, solutions and results to a problem in image processing involve a lot of trail and testing with huge set of sample images. Chlorella is a single-cell, freshwater green algae, and it consists of green plant pigments, chlorophyll, vitamins, minerals, and protein, fiber, and omega fatty acids. The size of the chlorella cells are 10–30 μm. Due to its photosynthetic process, it converts carbon dioxide into fresh oxygen. Automatic identification and classification of algal community are very difficult due to various factors such as change in size and shape with climatic changes, various growth periods, and the presence of other microbes. In this chapter, an elaborate analysis of artificial neural network concepts and convolutional neural network (CNN) of deep learning technique that automatically measure the algae growth through the image classification techniques from algae digital images using MATLAB is presented.

When medical images are processed by morphological operations, they provide substantial amount of utilizable information. The technological advancement in the field of image analysis and medical imaging domain acquiesces the understating of detection and diagnosis of disease to enhance the quality of medical treatment. Application of image processing in medical imaging field administers the development of processing nebulous, skeptical, reciprocal, superfluous information, and data for a vigorous structural attribute. To understand any image, the human and artificial astuteness system matches the features extracted from an image. Image enhancement is a decisive stage in image processing system. It intents at convalescencing the ocular data and the informational trait of wry images. After the acquisition of an image, if it is of poor quality, it requires enhancement. Various available techniques can be applied for enhancement; some are providing good results with limitation of computing time. A new intelligent algorithmic approach, based upon biologically inspired approaches, is suggested for image enhancement. In this ambience, this article describes about one of the most commonly used algorithms known as artificial bee colony algorithm, and its various types, used for image enhancement in different subdomains of medical imaging, are covered here.

The chapter gives an introduction to optimization based on evolutionary computational techniques and swarm intelligence. Evolutionary computational algorithms adopt the principles of biological evolution and use a population of solutions that evolves with every generation. The bio-inspired computing algorithms that mimic the behavior of swarms of birds and insects, referred collectively as swarm intelligence, are a subset of evolutionary algorithms. The behavior of swarms individually as well as collective behavior in a flock has been extensively studied and an insight into their integration with the optimization algorithm is given. The evolutionary optimization algorithms such as genetic algorithm, particle swarm optimization, ant colony optimization, bee colony optimization, cuckoo search, fish school search, firefly algorithm have been reviewed. The application of these algorithms to image processing has been outlined, and few case studies have been presented.

The proposed method is to analyze brain tumor to identify the grade of glioma from magnetic resonant image and histopathological images. The proposed work includes three phases. The first phase preprocesses the pathological images. This involves enhancement and contrast improvement of the images. Secondly, the processed histopathology image is subjected to feature extraction. Gaussian filters techniques and statistical feature extraction techniques are utilized for feature extraction. In the last phase, classifiers are developed to classify the low-grade and high-grade images based on extracted features. K-mean clustering network and SVD classifier are used for classification of low-grade/high-grade gliomas. MATLAB, a familiar tool, efficiently uses algorithms and techniques for identification of low-grade and high-grade glioma tumors.

Super-resolution (SR) reconstructs a high-resolution (HR) image from a set of low-resolution (LR) pictures and restores an HR video from a group of neighboring LR frames. Optimization tries to overcome the image acquisition limitations, the ill-posed nature of the SR problem, to facilitate content visualization and scene recognition. Particle swarm optimization (PSO) is a superb optimization algorithm used for all sorts of problems despite its tendency to be stuck in local minima. To handle ill-posedness, different PSO variants (hybrid versions) have been proposed trying to explore factors such as the initialization of the swarm, insertion of a constriction coefficient, mutation operators, and the use of an inertia weight. Hybridization involves combining two (or more) techniques wisely such that the resultant algorithm contains the good characteristics of both (or all) the methods. Interesting hybridization techniques include many local and global search approaches. Results for the SR reconstruction of still and video images are presented for the PSO and the HPSO algorithms.