Showing papers in "ICTACT Journal on Image and Video Processing in 2016"

Classification of cervical cancer cells in pap smear screening test

Abstract: Cervical cancer is second topmost cancers among women but also, it was a curable one. Regular smear test can discover the sign of precancerous cell and treated the patient according to the result. However sometimes the detection errors can be occurred by smear thickness, cell overlapping or by un-wanted particles in the smear and cytotechnologists faulty diagnosis. Therefore the reason automatic cancer detection was developed. This was help to increase cancer cell mindfulness, diagnosis accuracy with low cost. This detection process consists of some techniques of the image preprocessing that is segmentation and effective texture feature extraction with SVM classification. Then the Final Classification Results of this proposed technique was compared to the previous classification techniques of KNN and ANN and the result would be very useful to cytotechnologists for their further analysis.

A novel wrapping curvelet transformation based angular texture pattern (wctatp) extraction method for weed identification

Abstract: Apparently weed is a major menace in crop production as it competes with crop for nutrients, moisture, space and light which resulting in poor growth and development of the crop and finally yield. Yield loss accounts for even more than 70% when crops are frown under unweeded condition with severe weed infestation. Weed management is the most significant process in the agricultural applications to improve the crop productivity rate and reduce the herbicide application cost. Existing weed detection techniques does not yield better performance due to the complex background, illumination variation and crop and weed overlapping in the agricultural field image. Hence, there arises a need for the development of effective weed identification technique. To overcome this drawback, this paper proposes a novel Wrapping Curvelet Transformation Based Angular Texture Pattern Extraction Method (WCTATP) for weed identification. In our proposed work, Global Histogram Equalization (GHE) is used improve the quality of the image and Adaptive Median Filter (AMF) is used for filtering the impulse noise from the image. Plant image identification is performed using green pixel extraction and k-means clustering. Wrapping Curvelet transform is applied to the plant image. Feature extraction is performed to extract the angular texture pattern of the plant image. Particle Swarm Optimization (PSO) based Differential Evolution Feature Selection (DEFS) approach is applied to select the optimal features. Then, the selected features are learned and passed through an RVM based classifier to find out the weed. Edge detection and contouring is performed to identify the weed in the plant image. The Fuzzy rule-based approach is applied to detect the low, medium and high levels of the weed patchiness. From the experimental results, it is clearly observed that the accuracy of the proposed approach is higher than the existing Support Vector Machine (SVM) based approaches. The proposed approach achieves better performance in terms of accuracy.

Computer aided diagnosis for detection and stage identification of cervical cancer by using pap smear screening test images

Abstract: The majority of the women of the world were affected by the disease of cervical cancer. As a result of this disease, their death rate was increase as hasty level. Hence so many number of research people was focused this notion as their research interest and also they have done so many number of solutions for finding this cancer by using some image processing technique and achieved a good results only in advanced and high cost techniques of LBC, biopsy or Colposcopy test Images. Therefore the reason, the authors have chosen this problem and also did not only to find whether the patient is affected by a cancer or not. In addition to the patient was affected by this cancer means and also to identify which severity stage of this disease the patient could be live. Then this work has done in based on the images of low cost pap smear screening test by using various image processing techniques with the help of Computerized Image Processing Software Interactive Data Language (IDL-Image Processing Language). Thus the final reports would be very useful to the pathologists for further analysis.

Comparative study of edge based lsb matching steganography for color images

Abstract: Steganography is a very pivotal technique mainly used for covert transfer of information over a covert communication channel. This paper proposes a significant comparative study of the spatial LSB domain technique that focuses on sharper edges of the color as well as gray scale images for the purpose of data hiding and hides secret message first in sharper edge regions and then in smooth regions of the image. Message embedding depends on content of the image and message size. The experimental results illustrate that, for low embedding rate the method hides the message in sharp edges of cover image to get better stego image visualization quality. For high embedding rate, smooth regions and edges of the cover image are used for the purpose of data hiding. In this steganography method, color image and textured kind of image preserves better visual quality of stego image. The novelty of the comparative study is that, it helps to analyze the efficiency and performance of the method as it gives better results because it directly works on color images instead of converting to gray scale image

Image encryption in block-wise with multiple chaotic maps for permutation and diffusion

Abstract: This paper presents an efficient block-wise image encryption method based on multiple chaotic maps. The image is divided into four overlapping blocks and each block is permutated with Cat map and its parameters are controlled by Henon map using multiple keys. Due to overlapping division of blocks, it produces effect of double permutation in the middle portion of overlapped image in single permutation itself. For diffusion, the whole image is divided into four non-overlapping blocks and diffused with logistic map. Each block pixel values were completely modified in the diffusion process in order to avoid knownplaintext and chosen-plaintext attacks. For each division of blocks different keys were selected for both permutation and diffusion process in the proposed method. The simulation results of several statistical analysis shows that the proposed cryptosystem is efficient and highly secured.

Combining laplacian and sobel gradient for greater sharpening

Abstract: Generally medical images have narrow dynamic range of intensity levels and high noise. One such abdominal x-ray image has been taken for enhancement. This paper enhances the image by sharpening and by lime lighting more, the minute details of the bone structure present in the image. Actually the whole task has been accomplished with Laplacian filter to highlight fine details and with Sobel gradient to emphasize edges. To get the sharpened image, smoothed gradient image is used to mask the Laplacian image. Finally the dynamic range has been increased using gamma transformation.

An illumination invariant face recognition by enhanced contrast limited adaptive histogram equalization

Abstract: Face recognition system is gaining more importance in social networks and surveillance. The face recognition task is complex due to the variations in illumination, expression, occlusion, aging and pose. The illumination variations in image are due to changes in lighting conditions, poor illumination, low contrast or increased brightness. The variations in illumination adversely affect the quality of image and recognition accuracy. The illumination variations in face image have to be pre-processed prior to face recognition. The Contrast Limited Adaptive Histogram Equalization (CLAHE) is an image enhancement technique popular in enhancing medical images. The proposed work is to create illumination invariant face recognition system by enhancing Contrast Limited Adaptive Histogram Equalization technique. This method is termed as “Enhanced CLAHE”. The efficiency of Enhanced CLAHE is tested using Fuzzy K Nearest Neighbour classifier and fisher face subspace projection method. The face recognition accuracy percentage rate, Equal Error Rate and False Acceptance Rate at 1% are calculated. The performance of CLAHE and Enhanced CLAHE methods is compared. The efficiency of the Enhanced CLAHE method is tested with three public face databases AR, Yale and ORL. The Enhanced CLAHE has very high recognition accuracy percentage rate when compared to CLAHE.

Single frame super resolution of noncooperative iris images

Abstract: Image super-resolution, a process to enhance image resolution, has important applications in biometrics, satellite imaging, high definition television, medical imaging, etc. The long range captured iris identification systems often suffer from low resolution and meager focus of the captured iris images. These degrade the iris recognition performance. This paper proposes enhanced iterated back projection (EIBP) method to super resolute the long range captured iris polar images. The performance of proposed method is tested and analyzed on CASIA long range iris database by comparing peak signal to noise ratio (PSNR) and structural similarity index (SSIM) with state-of-the-art super resolution (SR) algorithms. It is further analyzed by increasing the up-sampling factor. Performance analysis shows that the proposed method is superior to state-of-the-art algorithms, the peak signal-tonoise ratio improved about 0.1-1.5 dB. The results demonstrate that the proposed method is well suited to super resolve the iris polar images captured at a long distance.

An efficient self-updating face recognition system for plastic surgery face

Abstract: Facial recognition system is fundamental a computer application for the automatic identification of a person through a digitized image or a video source. The major cause for the overall poor performance is related to the transformations in appearance of the user based on the aspects akin to ageing, beard growth, sun-tan etc. In order to overcome the above drawback, Self-update process has been developed in which, the system learns the biometric attributes of the user every time the user interacts with the system and the information gets updated automatically. The procedures of Plastic surgery yield a skilled and endurable means of enhancing the facial appearance by means of correcting the anomalies in the feature and then treating the facial skin with the aim of getting a youthful look. When plastic surgery is performed on an individual, the features of the face undergo reconstruction either locally or globally. But, the changes which are introduced new by plastic surgery remain hard to get modeled by the available face recognition systems and they deteriorate the performances of the face recognition algorithm. Hence the Facial plastic surgery produces changes in the facial features to larger extent and thereby creates a significant challenge to the face recognition system. This work introduces a fresh Multimodal Biometric approach making use of novel approaches to boost the rate of recognition and security. The proposed method consists of various processes like Face segmentation using Active Appearance Model (AAM), Face Normalization using Kernel Density Estimate/Point Distribution Model (KDE-PDM), Feature extraction using Local Gabor XOR Patterns (LGXP) and Classification using Independent Component Analysis (ICA). Efficient techniques have been used in each phase of the FRAS in order to obtain improved results.

Segmentation and quality analysis of long range captured iris image

Abstract: The iris segmentation plays a major role in an iris recognition system to increase the performance of the system. This paper proposes a novel method for segmentation of iris images to extract the iris part of long range captured eye image and an approach to select best iris frame from the iris polar image sequences by analyzing the quality of iris polar images. The quality of iris image is determined by the frequency components present in the iris polar images. The experiments are carried out on CASIA-long range captured iris image sequences. The proposed segmentation method is compared with Hough transform based segmentation and it has been determined that the proposed method gives higher accuracy for segmentation than Hough transform.

A comparative analysis of single and combination feature extraction techniques for detecting cervical cancer lesions

Abstract: Cervical cancer is the third most common form of cancer affecting women especially in third world countries. The predominant reason for such alarming rate of death is primarily due to lack of awareness and proper health care. As they say, prevention is better than cure, a better strategy has to be put in place to screen a large number of women so that an early diagnosis can help in saving their lives. One such strategy is to implement an automated system. For an automated system to function properly a proper set of features have to be extracted so that the cancer cell can be detected efficiently. In this paper we compare the performances of detecting a cancer cell using a single feature versus a combination feature set technique to see which will suit the automated system in terms of higher detection rate. For this each cell is segmented using multiscale morphological watershed segmentation technique and a series of features are extracted. This process is performed on 967 images and the data extracted is subjected to data mining techniques to determine which feature is best for which stage of cancer. The results thus obtained clearly show a higher percentage of success for combination feature set with 100% accurate detection rate.

Edge detection using multispectral thresholding

Abstract: Edge detection is a fundamental tool in image processing and computer vision, particularly in the areas of feature detection and extraction. Among various edge detection methods, Otsu method is one of the best optimal thresholding methods for general real world images with regard to uniformity and shape measures. In this paper, a multispectral thresholding algorithm using Otsu method is proposed to detect the edges in multispectral images. Natural, art and simulated images are considered for testing. Since the edges are well known in the simulated images, they are considered for performance evaluation. The results of proposed method, Edge Detection using MultiSpectral Thresholding (EDMST), are compared against the results of Canny Otsu, Improved Otsu, Median based Otsu and Improved Gray Image Otsu edge detection algorithms based on the human visual system, the number of edges and the number of pixels. The experimental results show that the proposed method achieves better performance and hence applied on Satellite images.

New human semen analysis system (casa) using microscopic image processing techniques

Abstract: Computer assisted semen analysis (CASA) helps the pathologist or fertility specialist to evaluate the human semen. Detail analysis of spermatozoa like morphology and motility is very important in the process of intrauterine insemination (IUI) or In-vitro fertilization (IVF) in infertile couple. The main objective for this new semen analysis is to provide a low cost solution to the pathologist and gynecologist for the routine raw semen analysis, finding the concentration of the semen with dynamic background removal and classify the spermatozoa type (grade) according to the motility and structural abnormality as per the WHO criteria. In this paper a new system , computer assisted semen analysis system is proposed in which hybrid approach is used to identify the moving object, scan line algorithm is applied for confirmation of the objects having tails, so that we can count the actual number of spermatozoa. For removal of background initially the dynamic background generation algorithm is proposed to create a background for background subtraction stage. The standard data set is created with 40× and 100× magnification from the different raw semen s. For testing the efficiency of proposed algorithm, same frames are applied to the existing algorithm. Another module of the system is focused on finding the motility and Type classification of individual spermatozoa.

De-blurring single photon emission computed tomography images using wavelet decomposition

Abstract: Single photon emission computed tomography imaging is a popular nuclear medicine imaging technique which generates images by detecting radiations emitted by radioactive isotopes injected in the human body. Scattering of these emitted radiations introduces blur in this type of images. This paper proposes an image processing technique to enhance cardiac single photon emission computed tomography images by reducing the blur in the image. The algorithm works in two main stages. In the first stage a maximum likelihood estimate of the point spread function and the true image is obtained. In the second stage Lucy Richardson algorithm is applied on the selected wavelet coefficients of the true image estimate. The significant contribution of this paper is that processing of images is done in the wavelet domain. Pre-filtering is also done as a sub stage to avoid unwanted ringing effects. Real cardiac images are used for the quantitative and qualitative evaluations of the algorithm. Blur metric, peak signal to noise ratio and Tenengrad criterion are used as quantitative measures. Comparison against other existing de-blurring algorithms is also done. The simulation results indicate that the proposed method effectively reduces blur present in the image.

A survey of retina based disease identification using blood vessel segmentation

Abstract: The colour retinal photography is one of the most essential features to identify the confirmation of various eye diseases. The iris is primary attribute to authenticate the human. This research work presents the survey and comparison of various blood vessel related feature identification, segmentation, extraction and enhancement methods. Additionally, this study is observed the various databases performance for storing the images and testing in minimal time. This paper is also provides the better performance techniques based on the survey.

Texture based land cover classification algorithm using gabor wavelet and anfis classifier

Abstract: Texture features play a predominant role in land cover classification of remotely sensed images. In this study, for extracting texture features from data intensive remotely sensed image, Gabor wavelet has been used. Gabor wavelet transform filters frequency components of an image through decomposition and produces useful features. For classification of fuzzy land cover patterns in the remotely sensed image, Adaptive Neuro Fuzzy Inference System (ANFIS) has been used. The strength of ANFIS classifier is that it combines the merits of fuzzy logic and neural network. Hence in this article, land cover classification of remotely sensed image has been performed using Gabor wavelet and ANFIS classifier. The classification accuracy of the classified image obtained is found to be 92.8%.

Character recognition of video subtitles

Abstract: An important task in content based video indexing is to extract text information from videos. The challenges involved in text extraction and recognition are variation of illumination on each video frame with text, the text present on the complex background and different font size of the text. Using various image processing algorithms like morphological operations, blob detection and histogram of oriented gradients the character recognition of video subtitles is implemented. Segmentation, feature extraction and classification are the major steps of character recognition. Several experimental results are shown to demonstrate the performance of the proposed algorithm.

A distributional approach for image retrieval using hotelling's t-square statistic

Abstract: This paper proposes a novel method, based on a statistical probability distributional approach, Hotelling’s T 2 statistic and Orthogonality test. If the input query image is structured, it is segmented into various regions according to its nature and structure. Otherwise, the image is treated as textured; and it is considered for the experiment as it is. The test statistic T 2 is applied on each region and compares it to the target image. If the test of hypothesis is accepted, it is inferred that the query and target images are same or similar. Otherwise, it is assumed that they belong to different groups. Moreover, the Eigen vectors are computed on each region, and the orthogonality test is employed to measure the angle between the two images. The obtained results outperform the existing methods.

Face recognition by embedding of dt-cwt coefficient using som and ensemble based classifier

Abstract: In real world, applications designing of a robust face recognition system have always been a big challenge. This paper presents an approach to face recognition using embedding of dual tree complex wavelet transform, Self-organizing map and ensemble of weak classifier. The DT-CWT is applied on the images to obtain dynamic and multi scale informational characterization of the face images. Thus, a multidimensional feature vector is formed by combining DT-CWT coefficients. Self Organizing Map (SOM) is further used in embedding the feature vector which also results in reduction of feature vector. Finally, ensembles of k-Nearest Neighbor (k-NN) weak classifiers are used for classification of recognition system. The proposed approach is tested on image of ORL database. The experiment shows an impressive recognition result that culminated during the testing.

Features based on neighborhood pixels density - a study and comparison

Abstract: In optical character recognition applications, the feature extraction method(s) used to recognize document images play an important role. The features are the properties of the pattern that can be statistical, structural and/or transforms or series expansion. The structural features are difficult to compute particularly from hand-printed images. The structure of the strokes present inside the hand-printed images can be estimated using statistical means. In this paper three features have been purposed, those are based on the distribution of B/W pixels on the neighborhood of a pixel in an image. We name these features as Spiral Neighbor Density, Layer Pixel Density and Ray Density. The recognition performance of these features has been compared with two more features Neighborhood Pixels Weight and Total Distances in Four Directions already studied in our work. We have used more than 20000 Devanagari handwritten character images for conducting experiments. The experiments are conducted with two classifiers i.e. PNN and k-NN.

A Comprehensive Study on Text Information Extraction from Natural Scene Images

Abstract: In Text Information Extraction (TIE) process, the text regions are localized and extracted from the images. It is an active research problem in computer vision applications. Diversity in text is due to the differences in size, style, orientation, alignment of text, low image contrast and complex backgrounds. The semantic information provided by an image can be used in different applications such as content based image retrieval, sign board identification etc. Text information extraction comprises of text image classification, text detection, localization, segmentation, enhancement and recognition. This paper contains a quick review on various text localization methods for localizing texts from natural scene images.

Evaluation of immunohistochemistry (ihc) marker her2 in breast cancer

Abstract: The paper discusses a novel approach involving algorithm implementation and hardware Devkit processing for estimating the extent of cancer in a breast tissue sample. The process aims at providing a reliable, repeatable, and fast method that could replace the traditional method of manual examination and estimation. Immunohistochemistry (IHC) and Fluorescence in situ Hybridization (FISH) are the two main methods used to detect the marker status in clinical practice. FISH is though more reliable than IHC, but IHC is widely used as it is cheaper, convenient to operate and conserve, the morphology is clear. The IHC markers are Estrogen receptor (ER, Progesterone receptor (PR), Human Epidermal Growth Factor (HER2) that give clear indications of the presence of cancer cells in the tissue sample. HER2 remains the most reliable marker for the detection of breast cancer. The Human Epidermal Growth Factor Receptor (HER2) markers are discussed in the paper, as it gives clear indications of the presence of cancer cells in the tissue sample. HER2 is identified based on the color and intensity of the cell membrane staining. The color and intensity is obviously based on the thresholding for classifying the cancerous cells into severity levels in terms of score to estimate the extent of spread of cancer in breast tissue. For HER2 evaluation, the percentage of staining is calculated in terms of ratio of stain pixel count to the total pixel count. The evaluation of HER2 is obtained through simulation software (MATLAB) using intensity based algorithm and same is run on embedded processor evaluation board Devkit 8500. The results are validated with doctors.

Real-time object detection in parallel through atomic transactions

Abstract: Object detection and tracking is important operation involved in embedded systems like video surveillance, Traffic monitoring, campus security system, machine vision applications and other areas. Detecting and tracking multiple objects in a video or image is challenging problem in machine vision and computer vision based embedded systems. Implementation of such an object detection and tracking systems are done in sequential way of processing and also it was implemented using hardware synthesize tools like verilog HDL with FPGA, achieves considerably lesser performance in speed and it does support lesser atomic transactions. There are many object detection and tracking algorithm were proposed and implemented, among them background subtraction is one of them. This paper proposes an implementation of detecting and tracking multiple objects based on background subtraction algorithm using java and .NET and also discuss about the architecture concept for object detection through atomic transactional, modern hardware synthesizes language called Bluespec.

Endocardium segmentation: an approach using local chan vese model with radial charge fitting curve

Abstract: Wall tracking and Endocardium segmentation in Echocardiography images is a prime requirement for the diagnosis of major cardiac diseases. To avoid manual procedures of wall tracing and to provide a quantitative aid in the diagnosis procedure to the cardiologist, a new approach based on Local Chan Vese Model is proposed. The Model is based on curve evolution, local statistical function and level set method, and the accuracy of result is based on contour placement. This initial contour is generated through Radial Charge Fitting curve which is an auto generated curve based on Gauss Law. It is found that the inclusion of Radial Charge Fitting Curve with Local Chan Vese Model provides an accurate Endocardium Segmentation. The proposed method is compared with the Local Chan Vese Model with manual initial contours. It is proved that the proposed Local Chan Vese Model with Radial Charge Fitting Curve is performing accurate Endocardium Segmentation with minimal iterations.

Analysis of abnormalities in common carotid artery images using multiwavelets

Abstract: According to the report given by World Health Organization, by 2030 almost 23.6 million people will die from cardiovascular diseases (CVD), mostly from heart disease and stroke. The main objective of this work is to develop a classifier for the diagnosis of abnormal Common Carotid Arteries (CCA). This paper proposes a new approach for the analysis of abnormalities in longitudinal B-mode ultrasound CCA images using multiwavelets. Analysis is done using HM and GHM multiwavelets at various levels of decomposition. Energy values of the coefficients of approximation, horizontal, vertical and diagonal details are calculated and plotted for different levels. Plots of energy values show high correlation with the abnormalities of CCA and offer the possibility of improved diagnosis of CVD. It is clear that the energy values can be used as an index of individual atherosclerosis and to develop a cost effective system for cardiovascular risk assessment at an early stage.

Automated detection of mitotic figures in breast cancer histopathology images using gabor features and deep neural networks

Abstract: The count of mitotic figures in Breast cancer histopathology slides is the most significant independent prognostic factor enabling determination of the proliferative activity of the tumor. In spite of the strict protocols followed, the mitotic counting activity suffers from subjectivity and considerable amount of observer variability despite being a laborious task. Interest in automated detection of mitotic figures has been rekindled with the advent of Whole Slide Scanners. Subsequently mitotic detection grand challenge contests have been held in recent years and several research methodologies developed by their participants. This paper proposes an efficient mitotic detection methodology for Hematoxylin and Eosin stained Breast cancer Histopathology Images using Gabor features and a Deep Belief NetworkDeep Neural Network architecture (DBN-DNN). The proposed method has been evaluated on breast histopathology images from the publicly available dataset from MITOS contest held at the ICPR 2012 conference. It contains 226 mitoses annotated on 35 HPFs by several pathologists and 15 testing HPFs, yielding an F-measure of 0.74. In addition the said methodology was also tested on 3 slides from the MITOSISATYPIA grand challenge held at the ICPR 2014 conference, an extension of MITOS containing 749 mitoses annotated on 1200 HPFs, by pathologists worldwide. This study has employed 3 slides (294 HPFs) from the MITOS-ATYPIA training dataset in its evaluation and the results showed F-measures 0.65, 0.72and 0.74 for each slide. The proposed method is fast and computationally simple yet its accuracy and specificity is comparable to the best winning methods of the aforementioned grand challenges.