Showing papers in "ICTACT Journal on Image and Video Processing in 2013"

A Medical Multi-Modality Image Fusion of CT/PET with PCA, DWT Methods

Abstract: This paper gives a view on the fusion of different modality images like PET and CT (Positron Emission Tomography & Computed Tomography) by two domain methods PCA and DWT methods. The spatial domain is PCA method, and another transformation domain method (DWT). In dwt decomposed coefficients of DWT (discrete wavelet transformation) are applied with the IDWT to get fused image information. Before that, choose a detailed part of decomposed coefficients by maximum selection and averaging the approximated part of DWT coefficients. In applying the PCA using eigen values and eigen vector of larger values as principal components and after to reconstruct using addition to these to get the fussed image of two modalities CT & PET. So that adds complimentary features of both anatomic, physiological and metabolic information in one image, provides better visual information in single image of patients in medical field. The analytic parameters like, MSE, PSNR, ENTROPY results are better enough to prove the methods each other.

Finger knuckle print recognition with sift and k-means algorithm

Abstract: In general, the identification and verification are done by passwords, pin number, etc., which is easily cracked by others. Biometrics is a powerful and unique tool based on the anatomical and behavioral characteristics of the human beings in order to prove their authentication. This paper proposes a novel recognition methodology of biometrics named as Finger Knuckle print (FKP). Hence this paper has focused on the extraction of features of Finger knuckle print using Scale Invariant Feature Transform (SIFT), and the key points are derived from FKP are clustered using K-Means Algorithm. The centroid of K-Means is stored in the database which is compared with the query FKP K-Means centroid value to prove the recognition and authentication. The comparison is based on the XOR operation. Hence this paper provides a novel recognition method to provide authentication. Results are performed on the PolyU FKP database to check the proposed FKP recognition method.

Multiwavelet transform in compression of medical images

Abstract: This paper analyses performance of multiwavelets a variant of wavelet transform on compression of medical images. To do so, two processes namely, transformation for decorrelation and encoding are done. In transformation stage medical images are subjected to multiwavelet transform using multiwavelets such as GeronimoHardin-Massopust, Chui Lian, Cardinal 2 Balanced (Cardbal2) and orthogonal symmetric/antsymmetric multiwavelet (SA4). Set partitioned Embedded Block Coder is used as a common platform for encoding the transformed coefficients. Peak Signal to noise ratio, bit rate and Structural Similarity Index are used as metrics for performance analysis. For experiment we have used various medical images such as Magnetic Resonance Image, Computed Tomography and X-ray images.

Fuzzy based image dimensionality reduction using shape primitives for efficient face recognition

Abstract: Today face recognition capability of the human visual system plays a significant role in day to day life due to numerous important applications for automatic face recognition. One of the problems with the recent image classification and recognition approaches are they have to extract features on the entire image and on the large grey level range of the image. The present paper overcomes this by deriving an approach that reduces the dimensionality of the image using Shape primitives and reducing the grey level range by using a fuzzy logic while preserving the significant attributes of the texture. The present paper proposed an Image Dimensionality Reduction using shape Primitives (IDRSP) model for efficient face recognition. Fuzzy logic is applied on IDRSP facial model to reduce the grey level range from 0 to 4. This makes the proposed fuzzy based IDRSP (FIDRSP) model suitable to Grey level co-occurrence matrices. The proposed FIDRSP model with GLCM features are compared with existing face recognition algorithm. The results indicate the efficacy of the proposed method.

An illumination invariant texture based face recognition

Abstract: Automatic face recognition remains an interesting but challenging computer vision open problem. Poor illumination is considered as one of the major issue, since illumination changes cause large variation in the facial features. To resolve this, illumination normalization preprocessing techniques are employed in this paper to enhance the face recognition rate. The methods such as Histogram Equalization (HE), Gamma Intensity Correction (GIC), Normalization chain and Modified Homomorphic Filtering (MHF) are used for preprocessing. Owing to great success, the texture features are commonly used for face recognition. But these features are severely affected by lighting changes. Hence texture based models Local Binary Pattern (LBP), Local Derivative Pattern (LDP), Local Texture Pattern (LTP) and Local Tetra Patterns (LTrPs) are experimented under different lighting conditions. In this paper, illumination invariant face recognition technique is developed based on the fusion of illumination preprocessing with local texture descriptors. The performance has been evaluated using YALE B and CMU-PIE databases containing more than 1500 images. The results demonstrate that MHF based normalization gives significant improvement in recognition rate for the face images with large illumination conditions.

Haar transform based estimation of chlorophyll and structure of the leaf

Abstract: In this paper, the health of a plant is estimated using various nondestructive Image Processing Techniques. Chlorophyll content was detected based on colour Image Processing. The Haar transform is applied to get size of leaf and the parameters.

Computer-aided detection of acinar shadows in chest radiographs

Abstract: Despite the technological advances in medical diagnosis, accurate detection of infectious tuberculosis (TB) still poses challenges due to complex image features and thus infectious TB continues to be a public health problem of global proportions. Currently, the detection of TB is mainly conducted visually by radiologists examining chest radiographs (CXRs). To reduce the backlog of CXR examination and provide more precise quantitative assessment, computer-aided detection (CAD) systems for potential lung lesions have been increasingly adopted and commercialized for clinical practice. CADs work as supporting tools to alert radiologists on suspected features that could have easily been neglected. In this paper, an effective CAD system aimed for acinar shadow regions detection in CXRs is proposed. This system exploits textural and photometric features analysis techniques which include local binary pattern (LBP), grey level co-occurrence matrix (GLCM) and histogram of oriented gradients (HOG) to analyze target regions in CXRs. Classification of acinar shadows using Adaboost is then deployed to verify the performance of a combination of these techniques. Comparative study in different image databases shows that the proposed CAD system delivers consistent high accuracy in detecting acinar shadows.

Performance analysis of secure medical image communication with digital signature and reversible watermarking

Abstract: Protection of Medical image contents becomes the important issue in computer network security. Digital Watermarking has becomes a promising technique for medical content authentication, it allows to embed relevant information with the image, which provides confidentiality, integrity and authentication by embedding Digital Signature (DS) with the Medical image. In this paper we focus on need for reversible watermarking, Medical Image Compression and security related problems in medical images, it comparing the performances of various lossless watermarking techniques for various medical image modalities like MRI (Magnetic Resonance Imaging), US (Ultrasonic), CT (Computed Tomography), Endoscopic and Angiographic images. Region of Interest (ROI) supporting lossless watermarking systems only considered for discussions. Performance of all lossless watermarking with Digital Signature is analyzed by means of four parameters Capacity Rate, PSNR (Peak Signal to Noise ratio), NPCR (Number of Pixel Change Rate) and Compression Ratio (CR). This Paper also introduces new mechanism for open network security for medical images. This lossless watermarking is responsible for recovering the altered medical image content of the system.

Implementation of artificial neural network for face recognition using gabor feature extraction

Abstract: Face detection and recognition is the first step for many applications in various fields such as identification and is used as a key to enter into the various electronic devices, video surveillance, and human computer interface and image database management. This paper focuses on feature extraction in an image using Gabor filter and the extracted image feature vector is then given as an input to the neural network. The neural network is trained with the input data. The Gabor wavelet concentrates on the important components of the face including eye, mouth, nose, cheeks. The main requirement of this technique is the threshold, which gives privileged sensitivity. The threshold values are the feature vectors taken from the faces. These feature vectors are given into the feed forward neural network to train the network. Using the feed forward neural network as a classifier, the recognized and unrecognized faces are classified. This classifier attains a higher face deduction rate. By training more input vectors the system proves to be effective. The effectiveness of the proposed method is demonstrated by the experimental results.

Haptic rendering of thin and soft objects

Abstract: In this work, we take up the problem of modeling deformations, both local and global, along with stable translations for radially-thin virtual objects especially like that of a wire or a spaghetti noodle, during haptic manipulation. To achieve this we recommend the use of mass-spring systems rather than geometric models like vertex based and free form deformation as they do not model the physics behind the interactions. Finite Equation Methods (FEMs) are also not chosen as they are computationally expensive for fast haptic interactions and force feedback. We have explored different types of distribution of masses within the volume of the object, in order to come up with a suitable distribution of masses and network of springs and dampers so that the simulations mimic the behavior of a real object. We also model the constraint forces like normal and frictional forces between the object and the plane on which it is kept. Further, we simulate the effect of a varying temperature distribution of the object and discuss how anisotropic deformation of an object may be effected. We demonstrate through experimentations that it is indeed possible to haptically interact with virtual soft objects.

Measurement of rnfl thickness using oct images for glaucoma detection

Abstract: The thickness of retinal nerve fiber layer (RNFL) is one of the pompous parameters for assessing the disease, Glaucoma. A substantial amount of vision can be lost before the patient becomes aware of any defect. Optical Coherence Tomography (OCT) provides enhanced depth and clarity of viewing tissues with high resolution compared with other medical imaging devices. It examines the living tissue non-invasively. This paper presents an automatic method to find the thickness of RNFL using OCT images. The proposed algorithm first extracts all the layers present in the OCT image by texture segmentation using Gabor filter method and an algorithm is then developed to segment the RNFL. The thickness measurement of RNFL is automatically displayed based on pixel calculation. The calculated thickness values are compared with the original values obtained from hospital. The result shows that the proposed algorithm is efficient in segmenting the region of interest without manual intervention. The effectiveness of the proposed method is proved statistically by the performance analysis.

An approach to reduce the storage requirement for biometric data in aadhar project

Abstract: AADHAR is an Indian Government Project to provide unique identification to each Citizen of India. The objective of the project is to collect all the personal details and the biometric traits from each individual. Biometric traits such as iris, face and fingerprint are being collected for authentication. All the information will be stored in a centralized data repository. Considering about the storage requirement for the biometric data of the entire population of India, approximately 20,218 TB of storage space will be required. Since 10 fingerprint data are stored, fingerprint details will take most of the space. In this paper, the storage requirement for the biometric data in the AADHAR project is analyzed and a method is proposed to reduce the storage by cropping the original biometric image before storing. This method can reduce the storage space of the biometric data drastically. All the measurements given in this paper are approximate only.

A Wavelet Transform Based Watermarking Algorithm for Protecting Copyrights of Digital Images

Abstract: This paper proposes an algorithm of Digital Watermarking based on Biorthogonal Wavelet Transform. Digital Watermarking is a technique to protect the copyright of the multimedia data. The position of the watermark can be detected without using the original image by utilizing the correlation between the neighbours of wave coefficient. The strength of Digital watermark is obtained according to the edge intensities resulting in good robust and Imperceptible. Results show that the proposed watermark algorithm is invisible and has good robustness against common image processing operations.

A combined approach using textural and geometrical features for face recognition

Abstract: Texture feature plays a predominant role in recognizing face images. However different persons can have similar texture features that may degrade the system performance. Hence in this paper, the problem of face similarity is addressed by proposing a solution which combines textural and geometrical features. An algorithm is proposed to combine these two features. Five texture descriptors and few geometrical features are considered to validate the proposed system. Performance evaluations of these features are carried out independently and jointly for three different issues such as expression variation, illumination variation and partial occlusion with objects. It is observed that the combination of textural and geometrical features enhance the accuracy of face recognition. Experimental results on Japanese Female Facial Expression (JAFFE) and ESSEX databases indicate that the texture descriptor Local Binary Pattern achieves better recognition accuracy for all the issues considered.

Digital color image encryption based on invertible matrix with secret sharing

Abstract: This paper explores the new approach to implement image encryption in digital color images. The self invertible matrix created from the original image is used as keys for the RGB to YCbCr transform and the secret sharing operations. The encryption process carried out by the four steps: pixel permutation, creating RGB matrix, RGB to YCbCr transform and the secret sharing. The quality of the encrypted images are tested with visual inspection and evaluated with different quality measures. The performance of the proposed method is also evaluated by various testing methods.

Analysis of tarsal tunnel syndrome using image correlation

Abstract: Tarsal tunnel syndrome (TTS), also known as posterior tibial neuralgia is a painful disorder of the foot. It is a medical condition arising due to the compression of the tibial nerve in the tarsal tunnel, resulting in numbness, parenthesis and muscle weakness in foot. A number of imaging methodologies such as ultrasound as well as MRI imaging has been used in the past in order to analyze the strain pattern of gastrocnemius tendon and aponeurosis from the surface of the skin without analyzing the internal tendons. The DIC code developed, computes the in-plane strain with a correlation function using pictures taken before and after stretching, using a CCD camera. The shift between the initial picture and subsequent one is evaluated by cross-correlation using FFT. This paper gives in detail description of the preprocessing steps necessary to extract Zone of Interest from the two images. The effects of stretching on the superficial components of the tibial nerve, the posterior tibial artery and vein, and the tibialis posterior, flexor digitorum longus and flexor hallucis longus tendons in the calf and foot are studied.

Radiance domain compositing for high dynamic range imaging

Abstract: High dynamic range imaging aims at creating an image with a range of intensity variations larger than the range supported by a camera sensor. Most commonly used methods combine multiple exposure low dynamic range (LDR) images, to obtain the high dynamic range (HDR) image. Available methods typically neglect the noise term while finding appropriate weighting functions to estimate the camera response function as well as the radiance map. We look at the HDR imaging problem in a denoising frame work and aim at reconstructing a low noise radiance map from noisy low dynamic range images, which is tone mapped to get the LDR equivalent of the HDR image. We propose a maximum aposteriori probability (MAP) based reconstruction of the HDR image using Gibb’s prior to model the radiance map, with total variation (TV) as the prior to avoid unnecessary smoothing of the radiance field. To make the computation with TV prior efficient, we extend the majorize-minimize method of upper bounding the total variation by a quadratic function to our case which has a nonlinear term arising from the camera response function. A theoretical justification for doing radiance domain denoising as opposed to image domain denoising is also provided.

Removal of impulsive noise using weighted fuzzy mean filter based on cloud model

Abstract: This paper proposes a weighted fuzzy mean filter based on cloud model and reports its performance in removing the impulsive noise from the digital image. In addition, the performance of the proposed weighted fuzzy mean filter is compared with already existing variants of median and switching filters using root mean square error, peak signal to noise ratio and quality index. Even though the image is corrupted by 90%, this weighted fuzzy mean filter is capable of recovering the original image with good detail preservation.

A statistical sharpness measure based multi focus image fusion using double density discrete wavelet transform

Abstract: Image fusion is the process of combining two or more images of the same scene to form the fused image retaining important features from each image with extended information content. There are two approaches to image fusion, namely Spatial Fusion and Transform fusion. Transform fusion uses transform for representing the source image at multi scale. Due to the compactness, orthogonality and directional information, the Discrete Wavelet Transforms and its undecimated version are used for image fusion. These transforms can be implemented using perfect reconstruction Finite Impulse Response filter banks which are either symmetric or orthogonal. To design filters to have both symmetric and orthogonal properties, the number of filters is increased to generate M-band transform. Double density Discrete Wavelet Transform is an example of M-band DWT and consists of one scaling and two wavelet filters. In this paper, an approach for DDWT based image fusion is designed using statistical property of wavelet filters in representing the sharpness and its performance is measured in terms of Root Mean Square Error, Peak to Signal Noise Ratio, Quality Index.

Codebook enhancement for vector quantization using minmax values for image compression

Abstract: Vector Quantization (VQ) is one of the Lossy Image Compression Techniques. Vector Quantization comprises of three different phases: Codebook Generation, Image Encoding and Image Decoding. The efficiency of VQ depends on the quality of the codebook. In this paper, we have proposed a novel idea for improving the quality of codebook. The codebook is populated with high detail blocks, with which the edges of the objects in an image can be preserved. The high detail blocks are identified based on the maximum and minimum pixel values of the blocks. K-Means clustering method is also used to improve the quality of the codebook. The quality (PSNR) of the reconstructed images by the proposed method is better when compared to that of few existing codebook generation techniques. Standard images are used to test the performance of the proposed method.

Stray dog detection in wired camera network

Abstract: Existing surveillance systems impose high level of security on humans but lacks attention on animals. Stray dogs could be used as an alternative to humans to carry explosive material. It is therefore imperative to ensure the detection of stray dogs for necessary corrective action. In this paper, a novel composite approach to detect the presence of stray dogs is proposed. The captured frame from the surveillance camera is initially pre-processed using Gaussian filter to remove noise. The foreground object of interest is extracted utilizing ViBe algorithm. Histogram of Oriented Gradients (HOG) algorithm is used as the shape descriptor which derives the shape and size information of the extracted foreground object. Finally, stray dogs are classified from humans using a polynomial Support Vector Machine (SVM) of order 3. The proposed composite approach is simulated in MATLAB and OpenCV. Further it is validated with real time video feeds taken from an existing surveillance system. From the results obtained, it is found that a classification accuracy of about 96% is achieved. This encourages the utilization of the proposed composite algorithm in real time surveillance systems.

A Strong Security Protocol against Fingerprint Database Attacks

Abstract: The Biometric data is subject to on-going changes and create a crucial problem in fingerprint database. To deal with this, a security protocol is proposed to protect the finger prints information from the prohibited users. Here, a security protocol is proposed to protect the finger prints information. The proposed system comprised of three phases namely, fingerprint reconstruction, feature extraction and development of trigon based security protocol. In fingerprint reconstruction, the different crack variance level finger prints images are reconstructed by the M-band Dual Tree Complex Wavelet Transform (DTCWT). After that features are extracted by binarization. A set of finger print images are utilized to evaluate the performance of security protocol and the result from this process guarantees the healthiness of the proposed trigon based security protocol. The implementation results show the effectiveness of proposed trigon based security protocol in protecting the finger print information and the achieved improvement in image reconstruction and the security process.

High accurate low complex face detection based on kl transform and ycbcr gaussian model

Abstract: This paper presents a skin color model for face detection based on YCbCr Gauss model and KL transform. The simple gauss model and the region model of the skin color are designed in both KL color space and YCbCr space according to clustering. Skin regions are segmented using optimal threshold value obtained from adaptive algorithm. The segmentation results are then used to eliminate likely skin region in the gauss-likelihood image. Different morphological processes are then used to eliminate noise from binary image. In order to locate the face, the obtained regions are grouped out with simple detection algorithms. The proposed algorithm works well for complex background and many faces.

Comparative analysis of ds and ids algorithms in super-spatial structure prediction for medical image sequences

Abstract: With the rapid growth of digital technology the demand to preserve raw image data for further processing is increasing. In medical industry the images are generally in the form of sequences which are much correlated. These images are very important and hence lossless image compression is needed to reproduce the original quality of the image without any loss of information. The correlation among the image sequences is exploited by interframe coding. Interframe coding includes Motion Estimation and Motion Compensation process supported by the Block Matching Algorithm. There are various block matching algorithms. The proposed method has taken Diamond Search and Inverse Diamond Search for comparison. The algorithms are used in Super-Spatial Structure Prediction to achieve high compression ratio. Results are compared in terms of compression ratio and search points to the prior arts.

Fpga implementation of rsepd technique based impulse noise removal

Abstract: In the process of signals transmission and acquisition, image signals might be corrupted by impulse noise. Generally, digital images are corrupted by impulse noises. These are short duration noises, which degrade an image and are randomly distributed over the image. An efficient FPGA implementation for removing impulse noise in an image is presented in this paper. Existing techniques use standard median filter. These existing approaches changes the pixel values of both noise less and noisy pixels, so image might be blurred in nature. To avoid the changes on noise less pixels, an efficient FPGA implementation of Simple Edge Preserved De-noising technique (SEPD) and Reduced Simple Edge Preserved De-noising technique (RSEPD) are presented in this paper. In this technique, noise detection and noise removal operations are performed. This VLSI design gives better image quality. For 10 percentage noise added image, the obtained PSNR value of the image is 31.68 while denoising it.

Generalization of rayleigh maximum likelihood despeckling filter using quadrilateral kernels

Abstract: Speckle noise is the most prevalent noise in clinical ultrasound images. It visibly looks like light and dark spots and deduce the pixel intensity as murkiest. Gazing at fetal ultrasound images, the impact of edge and local fine details are more palpable for obstetricians and gynecologists to carry out prenatal diagnosis of congenital heart disease. A robust despeckling filter has to be contrived to proficiently suppress speckle noise and simultaneously preserve the features. The proposed filter is the generalization of Rayleigh maximum likelihood filter by the exploitation of statistical tools as tuning parameters and use different shapes of quadrilateral kernels to estimate the noise free pixel from neighborhood. The performance of various filters namely Median, Kuwahura, Frost, Homogenous mask filter and Rayleigh maximum likelihood filter are compared with the proposed filter in terms PSNR and image profile. Comparatively the proposed filters surpass the conventional filters.

Road detection using morphological operations in a complex scenario

Abstract: Image classification is an important research area in computer vision. Organizing images into semantic categories can be extremely useful for searching and browsing through large collections of images. It is a challenging task in various application domains, including satellite image classification, syntactic pattern recognition, medical diagnosis, biometry, video surveillance, vehicle navigation, industrial visual inspection, robot navigation etc. There are different approaches for image classification and imbalanced data classification. This paper provides a review of different methods for classifying images and imbalanced data classification. This paper proposes a method for road detection and highlights the importance of the imbalanced data classification in detecting the road in a complex scenario.