• Journal of Korea Multimedia Society
• /
• v.23 no.8
• /
• pp.965-976
• /
• 2020

We present in this work a segmentation method of E. coli bacterial images generated via phase contrast microscopy using a deep learning based hybrid feature generation. Unlike conventional machine learning methods that use the hand-crafted features, we adopt the denoising autoencoder in order to generate a precise and accurate representation of the pixels. We first construct a hybrid vector that combines original image, difference of Gaussians and image gradients. The created hybrid features are then given to a deep autoencoder that learns the pixels' internal dependencies and the cells' shape and boundary information. The latent representations learned by the autoencoder are used as the inputs of a softmax classification layer and the direct outputs from the classifier represent the coarse segmentation mask. Finally, the classifier's outputs are used as prior information for a graph partitioning based fine segmentation. We demonstrate that the proposed hybrid vector representation manages to preserve the global shape and boundary information of the cells, allowing to retrieve the majority of the cellular patterns without the need of any post-processing.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.40 no.5
• /
• pp.365-370
• /
• 2003

In this paper, a new watermarking algorithm that based on wavelet packet transform is proposed for remote sensing images, which include many high frequency components. It applies watermark to the overall subband that includes the lowest frequency band. Watermark is embedded on original image after selecting the significant wavelet packet coefficients. For selection of significant coefficients which watermarks is embedded on, zerotree algorithm is applied to wavelet packet coefficients using CPSO (Coefficient Partitioning Scanning Order). From the experimental result, the proposed algorithm shows better invisibility and robustness performance compare with conventional watermarking methods. Especially, it demonstrates better robustness for high image compression in the remote images.

• Journal of the Korean Society of Visualization
• /
• v.14 no.2
• /
• pp.52-60
• /
• 2016

The Passive Auxiliary Feedwater System(PAFS) is one of the advanced safety concepts adopted in the Advanced Power Reactor Plus(APR+). To validate the operational performance of the PAFS, detailed understanding of a boiling heat transfer on horizontal tube outside is of great importance. Especially, in the mechanistic boiling heat transfer model, it is important to visualize the phenomena but there are some limitations with conventional experimental approaches. In the present study, we devised a heater based on the Flexible Printed Circuit Board (FPCB) for a more comprehensive visualization and subsequently, a digital image processing technique for the bubble motion measurement was established. Using the measurement technique, important parameters of the nucleate boiling are analyzed.

• Convergence Security Journal
• /
• v.6 no.2
• /
• pp.53-60
• /
• 2006

In the applications such as CAD or image processing, a variety of geometric objects are manipulated. A polygon in which all the edges are parallel to x- or y-axis is decomposed into simple rectangles for efficient handling. But, depending on the partitioning algorithms, the same region can be decomposed into a completely different set of rectangles in the number, size and shape of rectangles. So, it is necessary an algorithm that compares two sets of rectangles extracted from two scenes such as CAD or image to see if they represent the same region. This paper proposes an efficient algorithm that compares two sets of rectangles. The proposed algorithm is not only simpler than the algorithm based on sweeping method, but also reduces the number $O(n^2)$ of overlapped rectangles from the algorithm based on a balanced binary tree to O(nlogn).

• Journal of KIISE:Computer Systems and Theory
• /
• v.31 no.5_6
• /
• pp.321-327
• /
• 2004

SPIHT algorithm is a wavelet based fast and effective technique for image compression. It uses a list structure to store status information which is generated during set-partitioning of toro-tree. Usually, this requires lots of additional memory depending on how high the bit-rate is. Therefore, in this paper, we propose a new technique called MZC-BFS, which needs no additional memory when running SPIHT algorithm. It explicitly performs a breadth first search of the spatial-tree using peano-code and eliminates additional memory as it uses pre-status significant test for encoding and LSB bits of some coefficients for decoding respectively. This method yields nearly the same performance as SPIHT. This may be desirable in fast and simple hardware implementation and reduces the cost of production because no lists and additional memory are required.

• KSCI Review
• /
• v.2 no.1
• /
• pp.130-139
• /
• 1995

This paper suggests the method to shorten the search area by using edge detection and subimage partition. For the purpose reduce encoding time, The Domain areas are reduced 1/64 by partitioning original image to subimage, and classified them into edge area and shade area so that detect only the area in the same class. for achieving an encoding with good fidelity, tried to differ the search method as the threshold value of edge which is included in subimage, and compared the compression rate and fidelity when set the size of range block as $4{\times}4$ and $8{\times}8$.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.25 no.4B
• /
• pp.708-714
• /
• 2000

In this paper, an one-dimensional VLSI array for high speed processing of fractal image compression algorithm based the quad-tree partitioning method is proposed. First of all, the single assignment code algorithm is derived from the sequential Fisher's algorithm, and then the data dependence graph(DG) is obtained. The two-dimension array is designed by projecting this DG along the optimal direction and the one-dimensional VLSI array is designed by transforming the obtained two-dimensional array. The number of Input/Output pins in the designed one-dimensional array can be reduced and the architecture of process elements(PEs) can he simplified by sharing the input pins of range and domain blocks and internal arithmetic units of PEs. Also, the utilization of PEs can be increased by reusing PEs for operations to the each block-size. For fractal image compression of 512X512gray-scale image, the proposed array can be processed fastly about 67 times more than sequential algorithm. The operations of the proposed one-dimensional VLSI array are verified by the computer simulation.

• Journal of KIISE:Databases
• /
• v.31 no.1
• /
• pp.13-22
• /
• 2004

With the proliferation of multimedia data, there is an increasing need to support the indexing and retrieval of high-dimensional image data. Although there have been many efforts, the performance of existing multidimensional indexing methods is not satisfactory in high dimensions. Thus the dimensionality reduction and the approximate solution methods were tried to deal with the so-called dimensionality curse. But these methods are inevitably accompanied by the loss of precision of query results. Therefore, recently, the vector approximation-based methods such as the VA- file and the LPC-file were developed to preserve the precision of query results. However, the performance of the vector approximation-based methods depend largely on the size of the approximation file and they lose the advantages of the multidimensional indexing methods that prune much search space. In this paper, we propose a new index structure called the GC-tree for efficient similarity search in image databases. The GC-tree is based on a special subspace partitioning strategy which is optimized for clustered high-dimensional images. It adaptively partitions the data space based on a density function and dynamically constructs an index structure. The resultant index structure adapts well to the strongly clustered distribution of high-dimensional images.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.36S no.12
• /
• pp.41-50
• /
• 1999

In this paper, the new VLSI architecture of a reconstruction filter for morphological image segmentation is proposed. The filter, based on the $h_{max}$ operation, simplifies the interior of each region while preserving the boundary information. The proposed architecture adopts a partitioned memory structure and an efficient image scanning strategy to reduce the operations. The proposed memory partitioning scheme makes it possible that every data required for processing can be read from each memory at a time, resulting in parallel data processing. By the extended connectivity consideration, the operation is much decreased because more simplification is achieved in scanning stage. The selective raster scan strategy endows the satisfactory noise removal capability with negligible hardware complexity increase. The proposed architecture is designed using VHDL, and functional evaluation is performed by the CAD tool, Mentor. The experiment results show that the proposed architecture can simplify image profile with less than 18% operations of the conventional method.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.30B no.10
• /
• pp.13-26
• /
• 1993

Motion picture algorithms are realized on the multiprocessor system presented in the Study I. For the most efficient processing of the algorithms, pipelining and geometrical parallel processing methods are employed, and processing time, communication load and efficiency of each algorithm are compared. The performance of the implemented system is compared and analysed with reference to MPEG coding algorithm. Theoretical calculations and experimental results both shows that geometrical partitioning is a more suitable parallel processing algorithm for moving picture coding having the advantage of easy algorithm modification and expansion, and the overall efficiency is higher than pipelining.