• Journal of Broadcast Engineering
• /
• v.14 no.4
• /
• pp.397-408
• /
• 2009

In this paper, we present a non-static color histogram method to retrieve clothing images that are similar to a query clothing. Given clothing area, our method automatically extracts major colors by using the octree-based quantization approach[16]. Then, a color palette that is composed of the major colors is generated. The feature of each clothing, which can be either a query or a database clothing image, is represented as a color histogram based on its color palette. We define the match color bins between two possibly different color palettes, and unify the color palettes by merging or deleting some color bins if necessary. The similarity between two histograms is measured by using the weighted Euclidean distance between the match color bins, where the weight is derived from the frequency of each bin. We compare our method with previous histogram matching methods through experiments. Compared to HSV cumulative histogram-based approach, our method improves the retrieval precision by 13.7 % with less number of color bins.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.38A no.1
• /
• pp.1-9
• /
• 2013

This paper propose a human action recognition method that uses bag-of-features (BoF) based on CS-LBP (center-symmetric local binary pattern) and a spatial pyramid in addition to the random forest classifier. To construct the BoF, an image divided into dense regular grids and extract from each patch. A code word which is a visual vocabulary, is formed by k-means clustering of a random subset of patches. For enhanced action discrimination, local BoF histogram from three subdivided levels of a spatial pyramid is estimated, and a weighted BoF histogram is generated by concatenating the local histograms. For action classification, a random forest, which is an ensemble of decision trees, is built to model the distribution of each action class. The random forest combined with the weighted BoF histogram is successfully applied to Standford Action 40 including various human action images, and its classification performance is better than that of other methods. Furthermore, the proposed method allows action recognition to be performed in near real-time.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2009.01a
• /
• pp.396-399
• /
• 2009

In this paper, we present a method to define a color similarity between color images using Octree-based quantization and similar color integration. The proposed method defines major colors from each image using Octree-based quantization. Two color palettes to consist of major colors are compared based on Euclidean distance and similar color bins between palettes are matched. Multiple matched color bins are integrated and major colors are adjusted. Color histogram based on the color palette is constructed for each image and the difference between two histograms is computed by the weighted Euclidean distance between the matched color bins in consideration of the frequency of each bin. As an experiment to validate the usefulness, we discriminated the same clothing from CCD camera images based on the proposed color similarity analysis. We retrieved the same clothing images with the success rate of 88 % using only color analysis without texture analysis.

• Phonetics and Speech Sciences
• /
• v.9 no.2
• /
• pp.77-83
• /
• 2017

We propose a new method to reduce emotion recognition errors caused by variation in speaker characteristics and speech rate. Firstly, for reducing variation in speaker characteristics, we adjust features from a test speaker to fit the distribution of all training data by using the histogram equalization (HE) algorithm. Secondly, for dealing with variation in speech rate, we augment the training data with speech generated in various speech rates. In computer experiments using EMO-DB, KRN-DB and eNTERFACE-DB, the proposed method is shown to improve weighted accuracy relatively by 34.7%, 23.7% and 28.1%, respectively.

• Journal of Electrical Engineering and Technology
• /
• v.10 no.4
• /
• pp.1877-1885
• /
• 2015

This paper presents a new Nearest-Neighbors based weighted representation for images and weighted K-Nearest-Neighbors (WKNN) classifier to improve the precision of image classification using the Bag of Visual Words (BOVW) based models. Scale-invariant feature transform (SIFT) features are firstly extracted from images. Then, the K-means++ algorithm is adopted in place of the conventional K-means algorithm to generate a more effective visual dictionary. Furthermore, the histogram of visual words becomes more expressive by utilizing the proposed weighted vector quantization (WVQ). Finally, WKNN classifier is applied to enhance the properties of the classification task between images in which similar levels of background noise are present. Average precision and absolute change degree are calculated to assess the classification performance and the stability of K-means++ algorithm, respectively. Experimental results on three diverse datasets: Caltech-101, Caltech-256 and PASCAL VOC 2011 show that the proposed WVQ method and WKNN method further improve the performance of classification.

• Proceedings of the IEEK Conference
• /
• 2000.07a
• /
• pp.450-453
• /
• 2000

In this paper, an automated segmentation algorithm is proposed for MR brain images using T1-weighted, T2-weighted, and PD images complementarily. The proposed segmentation algorithm is composed of 3 steps. In the first step, cerebrum images are extracted by putting a cerebrum mask upon the three input images. In the second step, outstanding clusters that represent inner tissues of the cerebrum are chosen among 3-dimensional (3D) clusters. 3D clusters are determined by intersecting densely distributed parts of 2D histogram in the 3D space formed with three optimal scale images. Optimal scale image best describes the shape of densely distributed parts of pixels in 2D histogram. In the final step, cerebrum images are segmented using FCM algorithm with it’s initial centroid value as the outstanding cluster’s centroid value. The proposed segmentation algorithm complements the defect of FCM algorithm, being influenced upon initial centroid, by calculating cluster’s centroid accurately And also can get better segmentation results from the proposed segmentation algorithm with multi spectral analysis than the results of single spectral analysis.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.20 no.3
• /
• pp.647-653
• /
• 2016

Recently, the data image processing has been applied to a variety of fields including broadcasting, communication, computer graphics, medicine, and so on. Generally, the image data may develop the noise during their transmission. Therefore, the studies have been actively conducted to remove the noise on the image. There are diverse types of noise on the image including salt and pepper noise, AWGN, and mixed noise. Hence, the filter algorithm for the image recovery was proposed that salt and pepper noise was processed by histogram and spatial weighted values after defining the noise to lessen the impact of mixed noise added in the image, and AWGN was processed by the pixel information of local mask establishing the weighted values in this study. Regarding the processed results by applying Lena images which were corrupted by salt and pepper noise(P=50%) and AWGN(${\sigma}=10$), suggested algorithm showed the improvement by 7.06[dB], 10.90[dB], 5.97[dB] respectively compared with the existing CWMF, A-TMF, AWMF.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.15 no.1
• /
• pp.42-47
• /
• 2005

In this papcr, we address the 3D reconstruction of the indoor circumstance using what the data is extracted by a pall of image from Stereo Camera. Generally sucaking, there arc three methods to extract 3-Dimensional data using IR sensor, Laser sensor and Stereo camera sensor. The best is stereo camera sensor which can show a high performance at a reasonable price. We used 'Window Correlation Matching Method' to extract 3-Dimensional data in stereo image. We proposed new Method to reduce error data, said 'Histogram Weighted Hough Transform'. Owing to this mettled, we reduced error data in each stereo image. So reconstruction is well done. 3-Dimensional Reconstruction is accomplished by using the DirectX that is well known as 3D-Game development tool. We show that the stereo camera can be not only used to extract 3-dimensional data but also applied to reconstruct the 3-Dimensional circumstance. And we try to reduce the error data using various method.

• Journal of Multimedia Information System
• /
• v.2 no.4
• /
• pp.303-310
• /
• 2015

Vulnerable road users such as bike, motorcycle, small automobiles, and etc. are easily attacked or threatened with bigger vehicles than them. So this paper suggests a new approach two-wheelers detection system riding on people based on modified histogram of oriented gradients (HOGs) which is weighted by curvature and local correlation coefficient. This correlation coefficient between two variables, in which one is the person riding a bike and other is its background, can represent correlation relation. First, we extract edge vectors using the curvature of Gaussian and Histogram of Oriented Gradients (HOG) which includes gradient information and differential magnitude as cell based. And then, the value, which is calculated by the correlation coefficient between the area of each cell and one of bike, can be used as the weighting factor in process for normalizing the HOG cell. This paper applied the Adaboost algorithm to make a strong classification from weak classification. The experimental results validate the effectiveness of our proposed algorithm show higher than that of the traditional method and under challenging, such as various two-wheeler postures, complex background, and even conclusion.

• Journal of Biomedical Engineering Research
• /
• v.32 no.4
• /
• pp.336-344
• /
• 2011

This article describes a new recognition algorithm using correlation coefficient for intelligent wheelchair to avoid collision for elderly or disabled people. The correlation coefficient can be used to represent the relationship of two different areas. The algorithm has three steps: Firstly, we extract an edge vector using the Histogram of Oriented Gradients(HOG) which includes gradient information and unique magnitude for each cell. From this result, the correlation coefficients are calculated between one cell and others. Secondly, correlation coefficients are used as the weighting factors for normalizing the HOG cell. And finally, these features are used to classify or detect variable and complicated shapes of two wheelers using Adaboost algorithm. In this paper, we propose a new feature vectors which is calculated by weighted cell unit to classify with multiple view-based shapes: frontal, rear and side views($60^{\circ}$, $90^{\circ}$ and mixed angle). Our experimental results show that two wheeler detection system based on a proposed approach leads to a higher detection accuracy than the method using traditional features in a similar detection time.