• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2002년도 ITC-CSCC -1
• /
• pp.204-207
• /
• 2002

The current paper proposes a new multispectral image data compression algorithm that can efficiently reduce spatial and spectral redundancies by applying classified prediction, a Karhunen-Loeve transform (KLT), and the three-dimensional set partitioning in hierarchical trees (3-D SPIHT) algorithm In the wavelet transform (WT) domain. The classification is performed in the WT domain to exploit the interband classified dependency, while the resulting class information is used for the interband prediction. The residual image data on the prediction errors between the original image data and the predicted image data is decorrelated by a KLT. Finally, the 3D-SPIHT algorithm is used to encode the transformed coefficients listed in a descending order spatially and spectrally as a result of the WT and KLT. Simulation results showed that the reconstructed images after using the proposed algorithm exhibited a better quality and higher compression ratio than those using conventional algorithms.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제8권8호
• /
• pp.2895-2912
• /
• 2014

Content-based image retrieval has been the most important technique for managing huge amount of images. The fundamental yet highly challenging problem in this field is how to measure the content-level similarity based on the low-level image features. The primary difficulties lie in the great variance within images, e.g. background, illumination, viewpoint and pose. Intuitively, an ideal similarity measure should be able to adapt the data distribution, discover and highlight the content-level information, and be robust to those variances. Motivated by these observations, we in this paper propose a probabilistic similarity learning approach. We first model the distribution of low-level image features and derive the free energy kernel (FEK), i.e., similarity measure, based on the distribution. Then, we propose a learning approach for the derived kernel, under the criterion that the kernel outputs high similarity for those images sharing the same class labels and output low similarity for those without the same label. The advantages of the proposed approach, in comparison with previous approaches, are threefold. (1) With the ability inherited from probabilistic models, the similarity measure can well adapt to data distribution. (2) Benefitting from the content-level hidden variables within the probabilistic models, the similarity measure is able to capture content-level cues. (3) It fully exploits class label in the supervised learning procedure. The proposed approach is extensively evaluated on two well-known databases. It achieves highly competitive performance on most experiments, which validates its advantages.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제10권12호
• /
• pp.5643-5656
• /
• 2016

The detection accuracy of steganalysis depends on many factors, including the embedding algorithm, the payload size, the steganalysis feature space and the properties of the cover source. In practice, the cover source mismatch (CSM) problem has been recognized as the single most important factor negatively affecting the performance. To address this problem, we propose a new framework for blind, universal steganalysis which uses traditional steganalyst features. Firstly, cover images with the same statistical properties are searched from a reference image database as aided samples. The test image and its aided samples form a whole test set. Then, by assuming that most of the aided samples are innocent, we conduct outlier detection on the test set to judge the test image as cover or stego. In this way, the framework has removed the need for training. Hence, it does not suffer from cover source mismatch. Because it performs anomaly detection rather than classification, this method is totally unsupervised. The results in our study show that this framework works superior than one-class support vector machine and the outlier detector without considering the image retrieval process.

• 한국컴퓨터정보학회논문지
• /
• 제24권1호
• /
• pp.85-92
• /
• 2019

In this paper, we propose a efficient feature point extraction method that can solve the problem of performance degradation by introducing a preprocessing process when extracting feature points by utilizing the characteristics of 360-degree realistic media. 360-degree realistic media is composed of images produced by two or more cameras and this image combining process is accomplished by extracting feature points at the edges of each image and combining them into one image if they cover the same area. In this production process, however, the stitching process where images are combined into one piece can lead to the distortion of non-seamlessness. Since the realistic media of 4K-class image has higher resolution than that of a general image, the feature point extraction and matching process takes much more time than general media cases.

• 센서학회지
• /
• 제30권5호
• /
• pp.286-289
• /
• 2021

Image classification is one of the fundamental applications of computer vision. It enables a system to identify an object in an image. Recently, image classification applications have broadened their scope from computer applications to edge devices. The convolutional neural network (CNN) is the main class of deep learning neural networks that are widely used in computer tasks, and it delivers high accuracy. However, CNN algorithms use a large number of parameters and incur high computational costs, which hinder their implementation in edge hardware devices. To address this issue, this paper proposes a lightweight image classifier that provides good accuracy while using fewer parameters. The proposed image classifier diverts the input into three paths and utilizes different scales of receptive fields to extract more feature maps while using fewer parameters at the time of training. This results in the development of a model of small size. This model is tested on the CIFAR-10 dataset and achieves an accuracy of 90% using .26M parameters. This is better than the state-of-the-art models, and it can be implemented on edge devices.

• Journal of Information Processing Systems
• /
• 제17권5호
• /
• pp.947-959
• /
• 2021

Aiming at the problem of low accuracy in the water boundary automatic extraction of islands from GF-2 remote sensing image with high resolution in three bands, new water edges automatic extraction method in island based on GF-2 remote sensing images, genetic algorithm (GA) method, is proposed in this paper. Firstly, the GA-OTSU threshold segmentation algorithm based on the combination of GA and the maximal inter-class variance method (OTSU) was used to segment the island in GF-2 remote sensing image after pre-processing. Then, the morphological closed operation was used to fill in the holes in the segmented binary image, and the boundary was extracted by the Sobel edge detection operator to obtain the water edge. The experimental results showed that the proposed method was better than the contrast methods in both the segmentation performance and the accuracy of water boundary extraction in island from GF-2 remote sensing images.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제12권1호
• /
• pp.392-412
• /
• 2018

Due to the semantic gap problem across different modalities, automatically retrieval from multimedia information still faces a main challenge. It is desirable to provide an effective joint model to bridge the gap and organize the relationships between them. In this work, we develop a deep image annotation and classification by fusing multi-modal semantic topics (DAC_mmst) model, which has the capacity for finding visual and non-visual topics by jointly modeling the image and loosely related text for deep image annotation while simultaneously learning and predicting the class label. More specifically, DAC_mmst depends on a non-parametric Bayesian model for estimating the best number of visual topics that can perfectly explain the image. To evaluate the effectiveness of our proposed algorithm, we collect a real-world dataset to conduct various experiments. The experimental results show our proposed DAC_mmst performs favorably in perplexity, image annotation and classification accuracy, comparing to several state-of-the-art methods.

• Journal of Information Processing Systems
• /
• 제12권1호
• /
• pp.149-168
• /
• 2016

The Active Appearance Model (AAM) is a class of deformable models, which, in the segmentation process, integrates the priori knowledge on the shape and the texture and deformation of the structures studied. This model in its sequential form is computationally intensive and operates on large data sets. This paper presents another framework to implement the standard version of the AAM model. We suggest a distributed and parallel approach justified by the characteristics of the model and their potentialities. We introduce a schema for the representation of the overall model and we study of operations that can be parallelized. This approach is intended to exploit the benefits build in the area of advanced image processing.

• International Journal of Control, Automation, and Systems
• /
• 제6권6호
• /
• pp.828-835
• /
• 2008

We present a novel method for eye location by means of a two-level classifier scheme. Locating the eye by machine-inspection of an image or video is an important problem for Computer Vision and is of particular value to applications in biomedical imaging. Our method aims to overcome the significant challenge of an eye-location that is able to maintain high accuracy by disregarding highly variable changes in the environment. A first level of computational analysis processes this image context. This is followed by object detection by means of a two-class discrimination classifier(second algorithmic level).We have tested our eye location system using FERET and BioID database. We compare the performance of two-level classifier with that of non-level classifier, and found it's better performance.

• 센서학회지
• /
• 제8권2호
• /
• pp.124-132
• /
• 1999

본 논문에서는 에지 영역 보상을 이용한 원격 센싱된 인공위성 화상의 대역간 벡터양자화 기법을 제안하였다. 이 기법에서는 인공위성 화상데이타의 분광적 반사 특성에 따라 각 화소벡터를 분류한 후, 각 분류영역에 대하여 대역내 및 대역간 중복성을 각각 제거하기 위하여 분류영역별 대역내 벡터양자화 및 분류영역별 대역간 벡터양자화를 행한다. 에지영역의 경우에 주변블럭의 영역정보 및 양자화 된 기준대역의 화소값을 이용하여 에지영역을 보상한다. 그후, 대역간 중복성을 효과적으로 제거하기 위하여 보상된 영역정보를 이용하여 분류영역별 대역간 벡터양자화를 행한다. 실제 원격 센싱된 인공위성 화상데이타에 대한 실험을 통하여 제안한 기법의 부호화 효율이 기존의 기법에 비하여 우수함을 확인하였다.