• Journal of the Korea Computer Graphics Society
• /
• v.28 no.2
• /
• pp.21-28
• /
• 2022

ASTC is one of the standard texture formats supported in OpenGL ES 3.2 and Vulkan 1.0 (and later versions), and it has been increasingly used on mobile platforms (Android and iOS). ASTC's most important feature is the block size configuration, thereby providing a trade-off between compression quality and rates. With the higher number of textures, however, it is difficult to manually determine the optimal block sizes of each texture. To solve the problem, we present a new approach based on PSNR values to automatically determine the ASTC block size. A brute-force approach, which compresses a texture on all block sizes and compares the PSNR values of the compressed textures, can increase the compression time by up to 14 times. In contrast, our three-step approach minimizes the compression-time overhead. According to our experiments on a texture set including 64 various textures, our method determined the block sizes from 4×4 to 12×12 and reduced the size of compressed files by 68%.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.46 no.3
• /
• pp.57-63
• /
• 2009

The rapid growth of computer-based image database, necessity of a system that can manage an image information is increasing. This paper presents a region-based image retrieval method using the combination of color(autocorrelogram), texture(CWT moments) and shape(Hu invariant moments) features. As a color feature, a color autocorrelogram is chosen by extracting from the hue and saturation components of a color image(HSV). As a texture, shape and position feature are extracted from the value component. For efficient similarity confutation, the extracted features(color autocorrelogram, Hu invariant moments, and CWT moments) are combined and then precision and recall are measured. Experiment results for Corel and VisTex DBs show that the proposed image retrieval algorithm has 94.8% Precision, 90.7% recall and can successfully apply to image retrieval system.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.8
• /
• pp.3312-3327
• /
• 2020

Plant diseases are a significant yield and quality constraint for farmers around the world due to their severe impact on agricultural productivity. Such losses can have a substantial impact on the economy which causes a reduction in farmer's income and higher prices for consumers. Further, it may also result in a severe shortage of food ensuing violent hunger and starvation, especially, in less-developed countries where access to disease prevention methods is limited. This research presents an investigation of Directional Local Quinary Patterns (DLQP) as a feature descriptor for plants leaf disease detection and Support Vector Machine (SVM) as a classifier. The DLQP as a feature descriptor is specifically the first time being used for disease detection in horticulture. DLQP provides directional edge information attending the reference pixel with its neighboring pixel value by involving computation of their grey-level difference based on quinary value (-2, -1, 0, 1, 2) in 0°, 45°, 90°, and 135° directions of selected window of plant leaf image. To assess the robustness of DLQP as a texture descriptor we used a research-oriented Plant Village dataset of Tomato plant (3,900 leaf images) comprising of 6 diseased classes, Potato plant (1,526 leaf images) and Apple plant (2,600 leaf images) comprising of 3 diseased classes. The accuracies of 95.6%, 96.2% and 97.8% for the above-mentioned crops, respectively, were achieved which are higher in comparison with classification on the same dataset using other standard feature descriptors like Local Binary Pattern (LBP) and Local Ternary Patterns (LTP). Further, the effectiveness of the proposed method is proven by comparing it with existing algorithms for plant disease phenotyping.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.24 no.4
• /
• pp.382-391
• /
• 2021

Unsupervised domain adaptation often gives impressive solutions to handle domain shift of data. Most of current approaches assume that unlabeled target data to train is abundant. This assumption is not always true in practices. To tackle this issue, we propose a general solution to solve the domain gap minimization problem without any target data. Our method consists of two regularization steps. The first step is a pixel regularization by arbitrary style transfer. Recently, some methods bring style transfer algorithms to domain adaptation and domain generalization process. They use style transfer algorithms to remove texture bias in source domain data. We also use style transfer algorithms for removing texture bias, but our method depends on neither domain adaptation nor domain generalization paradigm. The second regularization step is a feature regularization by feature alignment. Adding a feature alignment loss term to the model loss, the model learns domain invariant representation more efficiently. We evaluate our regularization methods from several experiments both on small dataset and large dataset. From the experiments, we show that our model can learn domain invariant representation as much as unsupervised domain adaptation methods.

• Journal of the Korean earth science society
• /
• v.26 no.7
• /
• pp.683-690
• /
• 2005

As remote sensing imagery with high spatial resolution (e.g. pixel resolution of 1m or less) is used widely in the specific application domains, the requirements of advanced methods for this imagery are increasing. Among many applicable methods, the texture image analysis, which was characterized by the spatial distribution of the gray levels in a neighborhood, can be regarded as one useful method. In the texture image, we compared and analyzed different results according to various directions, kernel sizes, and parameter types for the GLCM algorithm. Then, we studied spatial feature characteristics within each result image. In addition, a template matching program which can search spatial patterns using template images selected from original and texture images was also embodied and applied. Probabilities were examined on the basis of the results. These results would anticipate effective applications for detecting and analyzing specific shaped geological or other complex features using high spatial resolution imagery.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.44 no.1
• /
• pp.10-16
• /
• 2007

In this paper, we propose a technique for retrieving images using spatial color correlation and texture characteristics based on local fourier transform. In order to retrieve images, two new descriptors are proposed. One is a color descriptor which represents spatial color correlation. The other is a descriptor combining the proposed color descriptor with texture descriptor. Since most of existing color descriptors including color correlogram which represent spatial color correlation considered just color distribution between neighborhood pixels, the structural information of neighborhood pixels is not considered. Therefore, a novel color descriptor which simultaneously represents spatial color distribution and structural information is proposed. The proposed color descriptor represents color distribution of Min-Max color pairs calculating color distance between center pixel and neighborhood pixels in a block with 3x3 size. Also, the structural information which indicates directional difference between minimum color and maximum color is simultaneously considered. Then new color descriptor(min-max color correlation descriptor, MMCCD) containing mean and variance values of each directional difference is generated. While the proposed color descriptor includes by far smaller feature vector over color correlogram, the proposed color descriptor improves 2.5 % ${\sim}$ 13.21% precision rate, compared with color correlogram. In addition, we propose a another descriptor which combines the proposed color descriptor and texture characteristics based on local fourier transform. The combined method reduces size of feature vector as well as shows improved results over existing methods.

• Journal of Broadcast Engineering
• /
• v.20 no.6
• /
• pp.912-920
• /
• 2015

Unlike general texture-based feature description algorithms, Locally Likely Arrangement Hashing (LLAH) algorithm describes a feature based on the geometric relationship between its neighbors. Thus, even in poor-textured scenes or large camera pose changes, it can successfully describe and track features and enables to implement augmented reality. This paper aims to optimize the performance of LLAH algorithm for tracking random dots (= features) with Gaussian noise. For this purpose, images with different number of features and magnitude of Gaussian noise are prepared. Then, the performance of LLAH algorithm according to the conditions: the number of neighbors, the type of geometric invariants, and the distance between features, is analyzed, and the optimal conditions are determined. With the optimal conditions, each feature could be matched and tracked in real-time with a matching rate of more than 80%.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.31B no.9
• /
• pp.101-111
• /
• 1994

The human face can provide a great deal of information in regard to his/her race, age, sex, personality, feeling, psychology, mental state, health condition and ect. If we pay a close attention to the aging process, we are able to find out that there are recognizable phenomena such as eyelid drooping, cheek drooping, forehead furrowing, hair falling-out, the hair becomes gray and etc. This paper proposes that the method to estimate the age by analyzing these feature components for the facial image. Ang we also introduce the method of facial image synthesis in accordance with the cange of age. The feature components according to the change of age can be obtainec by dividing the facial image into the 3-dimensional shape of a face and the texture of a face and then analyzing the principle component respectively using 3-dimensional model. We assume the age of the facial image by comparing the extracted feature component to the facial image and synthesize the resulted image by adding or subtracting the feature component to/from the facial image. As a resurt of this simulation, we have obtained the age changed ficial image of high quality.

• Journal of IKEEE
• /
• v.18 no.4
• /
• pp.536-543
• /
• 2014

Terrain classification is still a challenging issue in image processing, especially with high resolution satellite images. The well-known obstacles include low accuracy in the detection of targets, especially for the case of man-made structures, such as buildings and roads. In this paper, we present an efficient approach to classify and detect building footprints, foliage, grass and road from high resolution grayscale satellite images. Our contribution is to build a strong classifier using AdaBoost based on a combination of co-occurrence and Haar-like features. We expect that the inclusion of Harr-like feature improves the classification performance of the man-made structures, since Haar-like feature is extracted from corner features and rectangle features. Also, the AdaBoost algorithm selects only critical features and generates an extremely efficient classifier. Experimental result indicates that the classification accuracy of AdaBoost classifier is much higher than that of the conventional classifier using back propagation algorithm. Also, the inclusion of Harr-like feature significantly improves the classification accuracy. The accuracy of the proposed method is 98.4% for the target detection and 92.8% for the classification on high resolution satellite images.

• Journal of Korea Multimedia Society
• /
• v.16 no.9
• /
• pp.1057-1066
• /
• 2013

Personal verification using finger-knuckle-print(FKP) uses lines and creases at the finger-knuckle area, so the orientation information of texture is an important feature. In this paper, we propose an effective FKP verification method which extracts keypoints using SIFT algorithm and matches the keypoints by vector similarity. The vector is defined as a direction vector which connects a keypoint extracted from a query image and a corresponding keypoint extracted from a reference image. Since the direction vector is created by a pair of local keypoints, the direction vector itself represents only a local feature. However, it has an advantage of expanding a local feature to a global feature by comparing the vector similarity among vectors in two images. The experimental results show that the proposed method is superior to the previous methods based on orientation codes.