• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.29B no.10
• /
• pp.10-18
• /
• 1992

This paper considers and implementation of the memory system which provides simultaneous access to pq image points of block(p$\times$q), horizontal vector(1$\times$pq)and/vertical vector(pq$\times$1) in 2-dimension image array, where p and q are design parameters. This memory system consists of an address calculation circuit, address routing circuit, data routing circuit, module selection circuit and m memory modules where m>qp. The address calculation circuit computes pq addresses in parallel by using the difference of addresses among image points. Extra module assignment circuit is not used by improving module selection circuit with routhing circuit. By using Verilog-XL logic simulator, we verify the correctness of the memory system and estimate the performance. The implemented system provides simultaneous access to 16 image points and is 6 times faster than conventional memory system.

• Journal of Information Processing Systems
• /
• v.6 no.4
• /
• pp.435-452
• /
• 2010

This paper proposes a novel reversible data hiding scheme based on a Vector Quantization (VQ) codebook. The proposed scheme uses the principle component analysis (PCA) algorithm to sort the codebook and to find two similar codewords of an image block. According to the secret to be embedded and the difference between those two similar codewords, the original image block is transformed into a difference number table. Finally, this table is compressed by entropy coding and sent to the receiver. The experimental results demonstrate that the proposed scheme can achieve greater hiding capacity, about five bits per index, with an acceptable bit rate. At the receiver end, after the compressed code has been decoded, the image can be recovered to a VQ compressed image.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2019.05a
• /
• pp.401-403
• /
• 2019

In this paper, we propose a generative adversarial networks (GAN) based text-to-image generating method. In many natural language processing tasks, which word expressions are determined by their term frequency -inverse document frequency scores. Word2Vec is a type of neural network model that, in the case of an unlabeled corpus, produces a vector that expresses semantics for words in the corpus and an image is generated by GAN training according to the obtained vector. Thanks to the understanding of the word we can generate higher and more realistic images. Our GAN structure is based on deep convolution neural networks and pixel recurrent neural networks. Comparing the generated image with the real image, we get about 88% similarity on the Oxford-102 flowers dataset.

• Journal of the Korean Society for Industrial and Applied Mathematics
• /
• v.26 no.4
• /
• pp.263-279
• /
• 2022

This paper presents an automatic inspection of defects in semiconductor images. We devise a statistical method to find defects on homogeneous background from the observation that it has a log-normal distribution. If computer aided design (CAD) data is available, we use it to construct a signed distance function (SDF) and change the pixel values so that the average of pixel values along the level curve of the SDF is zero, so that the image has a homogeneous background. In the absence of CAD data, we devise a hybrid method consisting of a model-based algorithm and two neural networks. The model-based algorithm uses the first right singular vector to determine whether the image has a linear or complex structure. For an image with a linear structure, we remove the structure using the rank 1 approximation so that it has a homogeneous background. An image with a complex structure is inspected by two neural networks. We provide results of numerical experiments for the proposed methods.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.8 no.6
• /
• pp.1419-1427
• /
• 2007

In this paper, we propose an algorithm which implements a multi-sided side match finite-state vector quantization(MSMVQ). After extracting the edge information from an image and classifying the image into edge blocks or non-edge blocks, we construct an edge map. We subdivide edge blocks into sixteen classes using discrete cosine transform(DCT) AC coefficients. Based on edge map information, a state codebook is made from the master codebook, and side match calculation is done for two-sided or three-sided current block of image. For reducing transmitted bits, a decision is made whether or not to encode the non-edge blocks among the pre-coded blocks by using the master codebook. Also for reducing allocation bits of codeword indices to decoder, a variable length coder is used. Considering the comparison with side match finite-state vector quantization(SMVQ) and two-sided SMVQ(TSMVQ) algorithm about Zelda, Lenna, Bridge and Peppers image, the new algorithm shows better picture quality than SMVQ and TSMVQ respectively.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.35S no.6
• /
• pp.127-136
• /
• 1998

This paper proposes a new MVQMultistage Vector Quantizer) using a hybrid structure. While in a conventional MVQ, the quantizers of all stages perform the encoding procedure for input signals, we introduce a quantizer that performs selectively. The proposed quantizer with a hybrid structure is composed of a FSVQ(Finite-State Vector Quantizer) for the first stage and a ordinary VQ(Vector Quantizer) for the second stage. A input block is firstly encoded by the FSVQ of the first stage. If the Euclidean distortion measure between original signals and the codevector selected from the state codebook of the FSVQ is less than a prespecified value, only the FSVQ is used for image coding. Otherwise, both the FSVQ of the first stage and the ordinary VQ of the second stage are used for image coding. While the conventional MVQ has an advantage that can achieve low encoding complexity in comparison to the ordinary VQ, but has a disadvantage that is suboptimal with respect to the performance measure and can not achieve the bit rate reduction, the proposed method achieve not only the bit rate reduction but also the performance improvement.

• The KIPS Transactions:PartB
• /
• v.13B no.3 s.106
• /
• pp.309-314
• /
• 2006

Recently, multimedia DBMS is appeared to be the core technology of the information society to store, manage and retrieve multimedia data efficiently. In this paper, we propose a new method for content based-retrieval system using wavelet transform, energy value to extract automatically feature vector from image data, and suggest an effective retrieval technique through this method. Wavelet transform is widely used in image compression and digital signal analysis, and its coefficient values reflect image feature very well. The correlation in wavelet domain between query image data and the stored data in database is used to calculate similarity. In order to assess the image retrieval performance, a set of hundreds images are run. The method using standard derivation and mean value used for feature vector extraction are compared with that of our method based on energy value. For the simulation results, our energy value method was more effective than the one using standard derivation and mean value.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.53 no.2
• /
• pp.47-52
• /
• 2016

The used digital images in the smart device and small displayer has been a downscaled image. In this paper, the detection of the downscaling image forgery is proposed using the feature vector according to the pixel value's gradients. In the proposed algorithm, AR (Autoregressive) coefficients are computed from pixel value's gradients of the image. These coefficients as the feature vectors are used in the learning of a SVM (Support Vector Machine) classification for the downscaling image forgery detector. On the performance of the proposed algorithm, it is excellent at the downscaling 90% image forgery compare to MFR (Median Filter Residual) scheme that had the same 10-Dim. feature vectors and 686-Dim. SPAM (Subtractive Pixel Adjacency Matrix) scheme. In averaging filtering ($3{\times}3$) and median filtering ($3{\times}3$) images, it has a higher detection ratio. Especially, the measured performances of all items in averaging and median filtering ($3{\times}3$), AUC (Area Under Curve) by the sensitivity and 1-specificity is approached to 1. Thus, it is confirmed that the grade evaluation of the proposed algorithm is 'Excellent (A)'.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.1
• /
• pp.64-72
• /
• 1995

We propose a new image coding method which takes into account both statistical redundancy and perceptual irrelevancy of the DCT coefficients so as to provide a high quality of the reconstructed images with a reduced transmission bit rate First, a block of DCT coefficients are decomposed into 16 subvectors so as for a subvector to convey key information about one of the low-pass or the dirctional filtered images. Then, the most significant subvector is selected as the principal edge of the block and then vector quantized. After that, the residuals of the block are computed and then sequentially quantized through aforementioned procedure until the quantization distortion is smaller than the target distortion. The proposed scheme is good at encoding images with a variety of transmission bit rates, especially at very low bit rate coding. In addition, it is another benifit of the proposed scheme that an image can be quantized with a wide range of the transmission bit rates by simply adapting the stopping criterion of the sequential vector quantizer according to the target distortion of the reconstructed image.

• Journal of the Korea Society of Computer and Information
• /
• v.16 no.4
• /
• pp.1-13
• /
• 2011

In this paper, we propose a gesture recognition system to recognize motions using the shape information of MHI (Motion History Image). The system acquires MHI to provide information on motions from images with input and extracts the gradient images from such MHI for each X and Y coordinate. It extracts the shape information by applying the shape context to each gradient image and uses the extracted pattern information values as the feature values. It recognizes motions by learning and classifying the obtained feature values with a SVM (Support Vector Machine) classifier. The suggested system is able to recognize the motions for multiple people as well as to recognize the direction of movements by using the shape information of MHI. In addition, it shows a high ratio of recognition with a simple method to extract features.