David B. L. Bong

Detection of Neovascularization in Diabetic Retinopathy

Diabetic retinopathy has become an increasingly important cause of blindness. Nevertheless, vision loss can be prevented from early detection of diabetic retinopathy and monitor with regular examination. Common automatic detection of retinal abnormalities is for microaneurysms, hemorrhages, hard exudates, and cotton wool spot. However, there is a worse case of retinal abnormality, but not much research was done to detect it. It is neovascularization where new blood vessels grow due to extensive lack of oxygen in the retinal capillaries. This paper shows that various combination of techniques such as image normalization, compactness classifier, morphology-based operator, Gaussian filtering, and thresholding techniques were used in developing of neovascularization detection. A function matrix box was added in order to classify the neovascularization from natural blood vessel. A region-based neovascularization classification was attempted as a diagnostic accuracy. The developed method was tested on images from different database sources with varying quality and image resolution. It shows that specificity and sensitivity results were 89.4% and 63.9%, respectively. The proposed approach yield encouraging results for future development.

Performance analysis of single and combined bit-planes feature extraction for recognition in face expression database

Bit-planes for digital gray images have been used in many applications for special feature extraction. This paper presents analysis of single and combined bit-plane performance for face recognition. Novel approach of using bit-plane as input to Feedforward Neural Network is used. Comparison is done to analyze recognition rate of using single and combined bit-planes. Every single pixel in an 8-bits gray level digital image consists of 8 bits. Among these eight bit-planes, bit-planes 4, 5, 6 and 7 provide better recognition rates than bit-planes 0, 1, 2 and 3. Feed-Forward Neural Network is used in this paper for training and testing the bit-planes. The face database used for evaluation is CMU AMP Face Expression Database, which consists of 13 subjects, with 75 8-bits gray level images from each subject. Bit-planes 4, 5, 6 and 7 achieve over 88% accuracy after 15 image samples are trained. However, by using combination of these 4 bit-planes based on majority vote, false acceptance rate (FAR) can be reduced from over 80% to 26.2%. False rejection rate (FRR) of the combined approach is 2.2% and half total error rate (HTER) is 14.2%.

Blind image blur assessment by using valid reblur range and histogram shape difference

Abstract The presence of blur artifact is an annoyance to image viewers, and affects the perceived quality of the image. Telecommunication service providers and imaging product manufacturers are interested in this quality feedback for their process and product improvement. However, human-based quality feedback is tedious, expensive and has to be done in compliance with the standards for subjective evaluation such as the ITU-R BT. 500 standard. Thus, automatic assessment of images is proposed to overcome the difficulties in human-based evaluation. The automatic assessment is basically an objective estimation to predict the blur severity of an image. In this paper, a new model for blind estimation is proposed by using reblur algorithm to create reblur image and measure valid reblur range. Shape difference of local histograms is measured between the reblur and test images to produce the blur score. The proposed model is performed in the spatial domain without the need of data conversion or training. Experiment results show that the proposed model is highly correlated to human perception of blurriness, and performs better than other state-of-the-art blur metrics in the spatial domain.

Application of multilayer perceptron with backpropagation algorithm and regression analysis for long-term forecast of electricity demand: A comparison

Having an accurate forecast of future electricity usage is vital for utility companies to be able to provide adequate power supply to meet the demand. Two methods have been implemented to perform forecasting of electricity demand, namely, regression analysis (RA) and artificial neural networks (ANNs). We aim to compare these two methods in this paper using the mean absolute percentage error (MAPE) to measure the forecasting performance. The results show that ANNs are more effective than RA in long-term forecast. In addition to that, from our investigation into the effects of the inclusion of economic and social factors, such as population and gross domestic product (GDP), into the forecast, we conclude that the inclusion of economic and social factors do not improve the accuracy of the forecast of the chosen ANN model for electricity demand.

Objective blur assessment based on contraction errors of local contrast maps

Blur distortion appears in multimedia content due to acquisition, compression or transmission errors. In this paper, a method is proposed to predict blur severity based on the contraction errors of local contrast maps. The proposed method is developed from the observation that histogram distribution of natural image would contract according to the degree of blur distortion. In order to quantify the level of contraction, an efficient method of determining local contrast boundaries is used. The upper and lower bounds of local histogram distribution are defined for the original image, and outlying points beyond these bounds are used to form the local contrast map. For the corresponding patch of a blur image, the same values of upper and lower bounds are used and the local contrast map for the blur image could be produced. Total difference between local contrast maps of the original and blur images is the contraction errors which are used to derive the blur score. The proposed method has advantages in terms of computation efficiency, and is performed in the spatial domain without the need of data transformation, conversion or filtering. In addition, prior training is not required at all for the model. Implementation of the proposed method as a multimedia tool is useful for estimating blur severity in multimedia content. The performance of the proposed method is verified by using three different blur databases and compared to popular state-of-the-art methods. Experiment results show that the proposed blur metric has high correlation with human perception of blurriness.

Degradation of a dielectric barrier discharge plasma actuator

Dielectric barrier discharge is vulnerable to ion bombardment, radical species or ultraviolet radiations that can be emitted by plasma filaments in air under atmospheric pressure. In our experiments, traces of degradation on the actuator surface can be observed by naked eye after the discharge operation. The degradation could come from the non-uniformity of the electric field. Despite the degradation marks, some scratches due to the corona discharge process can be seen on the dielectric surface. The parametric study in this study reveals that the degradation on the actuator panel is subjected to a failure rate that increases with the cumulative time of plasma operation and the magnitude of supplied voltage. Besides, this study suggests that the severity of degradation can be lessened for a symmetric and larger gap design plasma actuator, since the concentration for the ion bombardment can be weakened at a particular discharge area.

Development of a fuzzy-logic-based Occurrence updating model for process FMEA

Risk priority number (RPN) ranking system is used to evaluate the risk level of failures, to rank failures, and to prioritize actions in traditional failure mode and effect analysis (FMEA). The RPN score is determined by multiplication of three input scores estimated by users, i.e., severity, occurrence, and detect. Even through this approach is simple, one of the problems is the difficulty in obtaining a good estimate of the severity, occurrence and detect ratings. Besides, it is a tedious job to update the ratings from time to time. In this paper, FMEA system with a proposed framework equipped with a fuzzy inference system based occurrence model to predict the occurrence score is proposed, and the fuzzy occurrence model is devised. In here, we propose a property for the fuzzy occurrence model, i.e., monotone output property. We try to derive the condition for the fuzzy occurrence model to be monotone such as that the derivative in non negative. From the derivation, a guideline on how input membership functions should be tuned is also provided. Simulation results are analyzed using real information collected from a semiconductor manufacturing environment.

Quality assessment of stereoscopic image by 3D structural similarity

Objective image quality assessment is proposed with the intention to substitute human-rated subjective evaluation by using computational method. Several types of two dimensional (2D) image quality metrics were proposed in the last decade to evaluate the quality of 2D images. When three dimensional (3D) or stereoscopic imaging gradually become popular in different areas of application, new objective quality assessments for 3D images had also been proposed. In this paper, a new method for assessing 3D image quality is proposed. This method is an improvement of the popular 2D structural similarity (SSIM) method with the addition of depth information to measure similarity between distorted and reference 3D images. The proposed method was tested on benchmark 3D image databases to gauge its performance. Experiment results show that predicted quality scores, as calculated from the proposed algorithm, are highly correlated with the corresponding subjective scores from manual evaluation. The significance and effectiveness of the proposed method were also evaluated by comparing its performance to other state-of-the-art 3D quality metrics.

An analysis of the effects of bit plane extraction in fingerprint recognition

This paper discusses about the effects of bit plane extraction in fingerprint recognition. An alternative approach to recognise a fingerprint from extracted bit plane is analysed in attempt to find the best bit plane used for recognition. An 8-bit greyscale fingerprint image is extracted into 8 different bit planes. Each bit plane of the images is then used as the input image for recognition. A fingerprint recognition algorithm using phase-only correlation (POC) is applied on the extracted bit planes. Based on the results of the analysis, the average recognition rate achieved in bit plane 7 is higher compared to the other bit planes. Three hundred samples of fingerprint images from FingerDOS are used for the experimental purposes.

Novel Face Recognition Approach Using Bit-Level Information and Dummy Blank Images in Feedforward Neural Network

Bit-level information is useful in image coding especially in image compression. A digital image is constructed by multilevel information of bits, called as bit-plane information. For an 8-bits gray level digital image, bit-plane extraction has ability to extract 8 layers of bit-plane information. Conventional neural network-based face recognition usually used gray images as training and testing data. This paper presents a novel method of using bit-level images as input to feedforward neural network. CMU AMP Face Expression Database is used in the experiments. Experiment result showed improvement in recognition rate, false acceptance rate (FAR), false rejection rate (FRR) and half total error rate (HTER) for the proposed method. Additional improvement is proposed by introducing dummy blank images which consist of plain 0 and 1 images in the neural network training set. Experiment result showed that the final proposed method of introducing dummy blank images improve FAR by 3.5%.