소형 유전자 알고리즘을 이용한 스테레오 정합

한규필,도경훈 東西大學校 1999 동서논문집 Vol.5 No.-

Genetic algorithm is an efficient search method using principles of natural selection and population genetic. In conventional genetic algorithms, however, the size of gene pool should be increased to tnsure a convergency. Therefore, many memory spaces and much computation time were needed. Also, sine child chromosomes were generated by chromosome crossover and gene mutation, the algorithms have a complex structure. Thus in this paper, a compact streo matching algorithm using a population-based incremental learning dased on probability vector is proposed to reduce these problems. The PBIL method is modified for matching environment. Since the proposed algorithm uses a probability vwctor and eliminates gene pool, chromosome crossover, and gene mutation, the matching algorithm is simple and the computation load is considerably reduced. Even if the characteristics of images are changed, stable outputs are obtained without the modifification of the matching algorithm.

영역정보와 유전 알고리즘을 이용한 스테레오 정합

도경훈,한규필 東西大學校 1999 동서논문집 Vol.5 No.-

In this paper, a new approach of stereo matching using genetic algorithm is presented to improve the depth reconstruction method. And the region information of the image is also used to treat various image sources. Genetic algorithms are efficient search methods based on principles of natural selection and population genetic. Genetic operators, such as crossover and mutation, are modified for matching environment. 2-D chromosome is used for the simple treatment of image signal. Since several constrains, defined as similarity and smoothness, are generally adopted to remove false matching candidates in stereo vision, the fitness function which identifies the fittest individual is formed by the constrains. In addition, the informed dispaeity generation and initial population based on intensity difference are applied to reduce the searching space of genetic operations. Experimental results show that the computation load is less than conventional matching methods with relaxation and relatively stable outputs are acquired.

Genetic Stereo Matching Algorithm with Locally Adaptive Chromosomes

Han, Kyu-Phil,Lee, Ky-Soo 金烏工科大學校 2000 論文集 Vol.21 No.-

A new stereo matching scheme using a genetic algorithm is presented to improve the depth reconstruction method of stereo vision systems. Genetic algorithms are efficient search methods based on principles of population genetics, i.e. mating, chromosome crossover, gene mutation, and natural selection. The proposed approach considers the matching environment as an optimization problem and finds the optimal solution by using an evolutionary strategy. Accordingly, genetic operators are adapted for the circumstances of stereo matching: 1) an individual is a disparity set. 2) a chromosome has a 2D structure for handling image signals efficiently, and 3) a fitness function is composed of certain constraints which are commonly used in stereo matching. Since the fitness function consists of intensity similarity and disparity smoothness, the matching and relaxation processes are considered at the same time in each generation. In order to acquire a disparity map consistent with the image appearance, a region of the input image, divided by zero-crossing points, is extracted and used in the determination of the chromosome shape. As a result, all chromosomes contain the external image form, and the disparity output coincides with the input image without any modification of the matching algorithm. In addition, an informed gene generation based on intensity difference is applied to reduce the searching space of the genetic operations.

Original Articles : Clinical Study of Urticaria Pigmentosa in Children

( Kyu Han Kim ),( Jun Kyu Oh ),( Phil Soo Ahn ),( Kyoung Chan Park ),( Sang Eun Moon ) 대한피부과학회 1995 Annals of Dermatology Vol.7 No.1

HDC를 이용한 고속 MSRCR 알고리즘

한규필(Kyu-Phil Han) 한국멀티미디어학회 2010 멀티미디어학회논문지 Vol.13 No.11

적응적 NMS와 관심점에 기반한 효율적인 특징추출 알고리즘

한규필 ( Kyu Phil Han ),문명주 ( Myung Ju Moon ) 한국화상학회 2012 한국화상학회지 Vol.18 No.3

본 논문은 Markless AR(Augmented Reality)에서 가장 많이 활용되는 SURF (Speeded Up Robust Features) 기법의 효율성을 높인 새로운 특징 추출 알고리즘을 제시한다. 기존의 SURF 알고리즘은 불변 특징점을 보장하기 위해 Scale Pyramid를 이용하며 고정된 크기의 블록단위 NMS(Non-Maximum Suppression)를 사용하기 때문에 많은 계산이 요구된다. 제안된 알고리즘은 관심점에서만 특징점을 구하는 적응적 NMS 기법을 제시하여, Scale Pyramid를 사용하지 않고 불변 특징점을 보장하면서 동시에 관심 영역에서 적응적 NMS를 수행함으로써 연산을 줄이는 효율적인 특징 추출 알고리즘을 개발하였다. A new effective feature extraction scheme, modifying SURF algorithm for markless AR, is presented in this paper. SURF algorithm used the scale pyramid to guarantee invariant features and the block based NMS of a fixed size window, so that the algorithm requires a lot of calculations. The proposed algorithm suggests an enhanced feature detection based on interest points instead of scale pyramid, in order to assure invariant features, and uses an adaptive NMS in the interested area for more effective calculations. As the result, the proposed algorithm preserves the performance of feature extraction similar to that of the conventional SURF, but obtains better results in computation loads and time.

RDWT 개념에 기반한 고속 방향성 영상보간 기법

한규필 ( Kyu Phil Han ) 한국화상학회 2015 한국화상학회지 Vol.21 No.4

본 논문에서는 가장 단순한 Haar 웨이블릿 변환에서 간축 연산이 생략된 RDWT 개념에 기반한 새로운 방향성 영상보간 방법을 제시한다. 대부분의 영상보간 기법들은 성능을 높이기 위해 경계의 방향정보를 이용하지만, 경계에 기초한 방향정보를 추출하는데 많은 연산량이 요구되는 단점이 있다. 그러므로, 최근 보간 기법에서는 경계정보를 빠르게 추출하기 위해 WT를 활용하는 기법이 이용되고 있다. 기존 WT은 기본적으로 스케일 함수를 사용하여 간축을 수행하기 때문에, 인접 2화소간의 주파수 정보가 독립적으로 추출된다. 이렇게 간축되어 제거된 주파수 성분은 완벽한 복원을 위해 불필요한 정보이지만, 웨이블릿을 이용한 보간에서는 외삽 연산을 제거하기 위해서 간축되지 않은 정보가 필요하다. 그러므로, 본 연구에서는 경계 방향을 빠르게 판단하고 내삽 연산만으로 보간을 수행하기 위해서 간축 연산을 이용하지 않는 비간축 DWT, 즉 RDWT의 개념을 최초로 이용하는 단순한 보간기법을 제안한다. 그리고 추가적으로 고해상도 보간 영상의 경계부분의 화질을 개선하기 위해서 고해상도 경계화소와 저해상도 영상에서 가장 닮은 화소 값과의 가중치 합으로 갱신하는 화소삽입 방법을 후처리 기법으로 제시한다. 실험결과 제안한 기법은 속도면에서 3차보간보다 빠르며, 화질은 기존기법들보다 우수한 결과를 획득할 수 있었다. This paper presents a new directional image interpolation algorithm based on a redundant discrete wavelet transform concept that the decimation operator is removed from the simplest Haar wavelet one. Most of interpolation methods use the directional information of edges for the better performance, but it requires much computation loads. Therefore, recent approaches are used to adopt wavelet transforms for the fast extraction of edge information. Since conventional WTs basically use the scale function performing the decimation, the frequency information of the two adjacent pixel is independently extracted. The skipped and eliminated frequency coefficients are not necessary for a perfect reconstruction but essential for the interpolation using wavelet to avoid an extrapolation process. Therefore, a simple interpolation scheme using the concept of undecimated DWT, or RDWT, is proposed for the fast decision of edge’s direction and for the only use of interleaving process. In addition, the pixel embedding scheme, that a high resolution pixel is updated with the weighted sum of the most similar pixel value at its lower resolution, is inserted to improve the edge pixels’ quality as a post-processing. Experimental results show that the computation time of the proposed algorithm was faster than that of bi-cubic interpolation and the performance of the proposed one were better than those of conventional ones.

이진화 영상분할기법과 적응적 융합 가중치를 이용한 광노출 보정기법

한규필(Kyu-Phil Han) 한국멀티미디어학회 2021 멀티미디어학회논문지 Vol.24 No.11

소형 유선자 알고리즘을 이용한 새로운 스테레오 정합

한규필(Kyu-Phil Han),배태면(Tae-Min Bae),권순규(Sun-Kyu Kwon),하영호(Yeong-Ho Ha) 대한전자공학회 1999 대한전자공학회 학술대회 Vol.1999 No.6

스테레오 정합을 위한 일반화된 유전 알고리즘

한규필(Kyu-Phil Han),김용석(Yong-Suk Kim),송근원(Kun-Woen Song),박영식(Young-Sik Park),하영호(Yeong-Ho Ha) 대한전자공학회 1997 대한전자공학회 학술대회 Vol.1997 No.11