• The Journal of the Acoustical Society of Korea
• /
• v.10 no.4
• /
• pp.27-37
• /
• 1991

This paper proposes the ESFVQ(energy subspace fuzzy vector quantization) that employs energy subspaces to reduce the quantizing distortion which is less than that of a fuzzy vector quatization. The ESFVQ is applied to a speaker adaptation method by which Korean words spoken by unknown speakers are recognized. By generating mapped codebooks with fuzzy histogram according to each energy subspace in the training procedure and by decoding a spoken word through the ESFVQ in the recognition proecedure, we attempt to improve the recognition rate. The performance of the ESFVQ is evaluated by measuring the quantizing distortion and the speaker adaptive recognition rate for DDD telephone area names uttered by 2 males and 1 female. The quatizing distortion of the ESFVQ is reduced by 22% than that of a vector quantization and by 5% than that of a fuzzy vector quantization, and the speaker adaptive recognition rate of the ESFVQ is increased by 26% than that without a speaker adaptation and by 11% than that of a vector quantization.

• The Journal of the Korea Contents Association
• /
• v.11 no.3
• /
• pp.18-25
• /
• 2011

In a manufacturing process of metal coins, surface defects of coins are manually detected. This paper describes an new method for detecting surface defects of metal coins on a moving conveyor belt using image processing. This method consists of multiple procedures: segmentation of a coin from the background, alignment of the coin to the model, projection of the aligned coin to the best eigen image space, and detection of defects by comparison of the projection error with an adaptive threshold. In these procedures, the alignement and the projection are newly developed in this paper for the detection of coin surface defects. For alignment, we use the histogram of the segmented coin, which converts two-dimensional image alignment to one-dimensional alignment. The projection reduces the intensity variation of the coin image caused by illumination and coin rotation change. For projection, we build multiple eigen image spaces and choose the best eigen space using estimated coin direction. Since each eigen space consists of a small number of eigen image vectors, we can implement the projection in real- time.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.42 no.6
• /
• pp.121-132
• /
• 2005

In the formation of images in a camera, the exposure time is appropriately adjusted to obtain a good image. Hence for a successful alignment of a sequence of images to the same scene, it is required to compensate the different exposure times. If we have no knowledge regarding the exposure time, then we should develop an algorithm that can compensate an image with respect to a reference image without using any camera formation models. In this paper, an exposure compensation is performed by designing predictors based on the conditional expectation between the reference and input images. Further, an adaptive predictor design is conducted to manage the irregular exposure or histogram problem. In order to alleviate the blocking artifact and the overfitting problems in the adaptive scheme, a smoothing technique, which uses the pixels of the adjacent blocks, is proposed. We successfully conducted the exposure compensation using real images obtained from digital cameras and the transmission electron microscopy.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.7
• /
• pp.201-208
• /
• 2014

The performance of a landmine detection system depends on consistent extractions of the features of landmines. Since landmines have diverse sizes, it is critical to select an appropriate window size to represent the landmine region consistently. Conventional detection systems are incapable of extracting consistent landmine features because they employ fixed window sizes. This paper proposes a window size selection method according to the size of a landmine. The proposed method selects an appropriate window size based on the type of a landmine estimated from the response signal of the system. Data on various types of soils and landmines were generated from a simulation program to evaluate the performance of the proposed method. The results verified that the proposed method, which employs an adaptive window size, yields a better landmine detection rate than the conventional methods, which employ fixed window sizes.

• Journal of Korea Multimedia Society
• /
• v.18 no.11
• /
• pp.1289-1301
• /
• 2015

Pedestrian detection (PD) is an essential task in various applications and sliding window-based methods utilizing HOG (Histogram of Oriented Gradients) or HOG-like descriptors have been shown to be very effective for accurate PD. However, due to exhaustive search across images, PD methods based on sliding window usually require heavy computational time. In this paper, we propose a real-time PD method for embedded visual surveillance with fixed backgrounds. The proposed PD method employs HOG descriptors as many PD methods does, but utilizes selective search so that it can save processing time significantly. The proposed selective search is guided by restricting searching to candidate regions extracted from Adaptive Gaussian Mixture Model (AGMM)-based background subtraction technique. Moreover, approximate computation of HOG descriptor and implementation in fixed-point arithmetic mode contributes to reduction of processing time further. Possible accuracy degradation due to approximate computation is compensated by applying an appropriate one among three offline trained SVM classifiers according to sizes of candidate regions. The experimental results show that the proposed PD method significantly improves processing speed without noticeable accuracy degradation compared to the original HOG-based PD and HOG with cascade SVM so that it is a suitable real-time PD implementation for embedded surveillance systems.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.51 no.8
• /
• pp.360-367
• /
• 2002

In this paper, genetic algorithm based adaptive image enhancement filtering scheme is proposed and Implemented on FPGA board. Conventional filtering methods require a priori noise information for image enhancement. In general, if a priori information of noise is not available, heuristic intuition or time consuming recursive calculations are required for image enhancement. Contrary to the conventional filtering methods, the proposed filter system can find optimal combination of filters as well as their sequent order and parameter values adaptively to unknown noise types using structured genetic algorithms. The proposed image enhancement filter system is mainly composed of two blocks. The first block consists of genetic algorithm part and fitness evaluation part. And the second block consists of four types of filters. The first block (genetic algorithms and fitness evaluation blocks) is implemented on host computer using C code, and the second block is implemented on re-configurabe FPGA board. For gray scale control, smoothing and deblurring, four types of filters(median filter, histogram equalization filter, local enhancement filter, and 2D FIR filter) are implemented on FPGA. For evaluation, three types of noises are used and experimental results show that the Proposed scheme can generate optimal set of filters adaptively without a pioi noise information.

• Journal of Korea Multimedia Society
• /
• v.15 no.7
• /
• pp.837-846
• /
• 2012

This paper presents a watermarking scheme based on the perspective distance for the secure mash-up service. The proposed scheme embeds the watermark of the location information of satellite image and the user information using edge color histogram, which is dissimilar to general digital image. Therefore, this scheme can trace the illegal distributor and can protect private information of user through the watermarking scheme that is adaptive to satellite image. Experimental results verified that our scheme has the invisibility and also the robustness against geometric attacks of rotation and translation.

• Journal of Biosystems Engineering
• /
• v.32 no.4
• /
• pp.256-262
• /
• 2007

Image processing systems have been used to measure the plant parameters such as size, shape and structure of plants. There are yet some limited applications for evaluating plant colors due to illumination conditions. This study was focused to present adaptive methods to analyze plant leaf color regardless of illumination conditions. Color patches attached on the calibration bars were selected to represent leaf colors of lettuces and to test a possibility of health monitoring of lettuces. Repeatability of assigning leaf colors to color patches was investigated by two-tailed t-test for paired comparison. It resulted that there were no differences of assignment histogram between two images of one lettuce that were acquired at different light conditions. It supported that use of the calibration bars proposed for leaf color analysis provided color constancy, which was one of the most important issues in a video color analysis. A health discrimination equation was developed to classify lettuces into one of two classes, SOUND group and POOR group, using the machine vision. The classification accuracy of the developed health discrimination equation was 80.8%, compared to farmers' decision. This study could provide a feasible method to develop a standard color chart for evaluating leaf colors of plants and plant health monitoring system using the machine vision.

• International Journal of Computer Science & Network Security
• /
• v.24 no.1
• /
• pp.119-126
• /
• 2024

Disease caused by the coronavirus (COVID-19) is sweeping the globe. There are numerous methods for identifying this disease using a chest imaging. Computerized Tomography (CT) chest scans are used in this study to detect COVID-19 disease using a pretrain Convolutional Neural Network (CNN) ResNet50. This model is based on image dataset taken from two hospitals and used to identify Covid-19 illnesses. The pre-train CNN (ResNet50) architecture was used for feature extraction, and then fully connected layers were used for classification, yielding 97%, 96%, 96%, 96% for accuracy, precision, recall, and F1-score, respectively. When combining the feature extraction techniques with the Back Propagation Neural Network (BPNN), it produced accuracy, precision, recall, and F1-scores of 92.5%, 83%, 92%, and 87.3%. In our suggested approach, we use a preprocessing phase to improve accuracy. The image was enhanced using the Contrast Limited Adaptive Histogram Equalization (CLAHE) algorithm, which was followed by cropping the image before feature extraction with ResNet50. Finally, a fully connected layer was added for classification, with results of 99.1%, 98.7%, 99%, 98.8% in terms of accuracy, precision, recall, and F1-score.

• Journal of radiological science and technology
• /
• v.36 no.3
• /
• pp.227-235
• /
• 2013

In adaptive radiotherapy(ART), generated composite dose of surrounding normal tissue on overall treatment course which is using deformable image registration from multistage images. Also, compared with doses summed by each treatment plan and clinical significance is considered. From the first of May, 2011 to the last of July, 2012. Patients who were given treatment and had the head and neck cancer with 3-dimension conformal radiotherapy or intensity modulated radiotherapy, those who were carried out adaptive radiotherapy cause of tumor shrinkage and weight loss. Generated composite dose of surrounding normal tissue using deformable image registration was been possible, statistically significant difference was showed to mandible($48.95{\pm}3.89$ vs $49.10{\pm}3.55$ Gy), oral cavity($36.93{\pm}4.03$ vs $38.97{\pm}5.08$ Gy), parotid gland($35.71{\pm}6.22$ vs $36.12{\pm}6.70$ Gy) and temporomandibular joint($18.41{\pm}9.60$ vs $20.13{\pm}10.42$ Gy) compared with doses summed by each treatment plan. The results of this study show significant difference between composite dose by deformable image registration and doses summed by each treatment plan, composite dose by deformable image registration may generate more exact evaluation to surrounding normal tissue in adaptive radiotherapy.