• Proceedings of the Korea Multimedia Society Conference
• /
• 2012.05a
• /
• pp.406-409
• /
• 2012

컬러 영상의 흑백화 기술은 영상 처리를 이용하는 다양한 분야에서 널리 활용되고 있는 기술이다. 하지만 기존의 일반적으로 사용되는 흑백화 기술은 컬러 영상에서의 색차 정보를 잃어버리는 문제점을 가지고 있다. 이러한 색차 정보 손실의 단점을 개선하기 위해 여러 방법들이 제안되었지만 대부분의 경우 최적의 결과를 얻기 위해서는 특정한 파라미터가 필요하다. 본 논문에서는 이러한 문제점을 해결하기 위해 영상의 Saliency map을 이용한 자동 흑백화 기술을 제안한다. 제안한 방법은 영상의 중요 정보를 통해 획득된 대표색상 정보를 이용해 기존의 방법과는 달리 특별한 파라미터의 입력이 필요하지 않는 장점을 가지고 있다. 실험결과는 제안한 방법이 기존 방법에 비해 매우 효율적임을 보여준다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2006.06a
• /
• pp.97-99
• /
• 2006

우리는 관절체 캐릭터에 대해 시간에 따른 인식의 정도를 측정하는 모션 특징점 맵을 제안한다. 모션 특징점 맵은 이미지 특징점 맵에서 사용된 가우시안 거리 방법을 응용하여 계산할 수 있다. 관절 계층 구조에서의 모션과 시각적 인지간의 관계를 고려하여, 우리는 관절 동선 모션 특징점 맵과 관절 구동 모션 특징점 맵의 두 가지 모션 특징점 맵을 정의하였다. 정의한 두 가지 모션 특징점 맵을 사용하여 한 프레임에서의 모션 특징점 맵 또한 계산할 수 있다. 계산된 모션 특징점 맵은 모션의 시놉시스 생성, 정운동학 연산량의 축소, 자동적 카메라 동선 생성 등 여러 가지 응용 분야에 적용할 수 있다. 실험을 통하여 우리는 모션에 대한 인식 기반적인 접근을 통해 모션의 질적인 향상은 물론 계산적인 퍼포먼스의 향상에도 많은 기여를 할 수 있음을 알 수 있었다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2012.07a
• /
• pp.392-393
• /
• 2012

영상에서의 화질 평가는 학술적으로나 산업적으로나 중요한 문제이지만 지금까지 주로 쓰여온 PSNR 과 같은 방법은 실제 사람이 인식하는 평가에 잘 부합하지 못하는 문제점이 있었다. 본 논문에서는 모든 밝기 영역에서 영상을 평가할 수 있는 HDR-VDP2 와 주목도 지도를 결합하여 범용적으로 사용이 가능하며 성능도 뛰어난 화질 측정 방법을 제시하고, 다양한 주목도 지도에 대하여 그 성능과 화질 평가 성능 사이의 관계를 살펴본다. 구현에는 주목도 지도를 가중치로 사용함으로써 간단하게 더 좋은 성능의 화질 평가 시스템을 만들었고 이를 실험으로 보였다. 또한 주목도 지도의 성능과 화질 평가 시스템의 성능 사이에는 약한 양의 상관관계가 있는 것으로 나타났는데 주목도 지도와 함께 구조적 특징점들의 정보를 성능 평가 시스템에 포함시키면 더 좋은 결과를 얻을 것으로 기대된다.

• Journal of KIISE
• /
• v.44 no.12
• /
• pp.1282-1289
• /
• 2017

Since the study of deep convolutional neural network became prevalent, one of the important discoveries is that a feature map from a convolutional network can be extracted before going into the fully connected layer and can be used as a saliency map for object detection. Furthermore, the model can use features from each different layer for accurate object detection: the features from different layers can have different properties. As the model goes deeper, it has many latent skip connections and feature maps to elaborate object detection. Although there are many intermediate layers that we can use for semantic segmentation through skip connection, still the characteristics of each skip connection and the best skip connection for this task are uncertain. Therefore, in this study, we exhaustively research skip connections of state-of-the-art deep convolutional networks and investigate the characteristics of the features from each intermediate layer. In addition, this study would suggest how to use a recent deep neural network model for semantic segmentation and it would therefore become a cornerstone for later studies with the state-of-the-art network models.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.12 no.6
• /
• pp.629-635
• /
• 2019

In Deep Learning models derivative is implemented by error back-propagation which enables the model to learn the error and update parameters. It can find the global (or local) optimal points of parameters even in the complex models taking advantage of a huge improvement in computing power. However, deliberately generated data points can 'fool' models and degrade the performance such as prediction accuracy. Not only these adversarial examples reduce the performance but also these examples are not easily detectable with human's eyes. In this work, we propose the method to detect adversarial datasets with random noise addition. We exploit the fact that when random noise is added, prediction accuracy of non-adversarial dataset remains almost unchanged, but that of adversarial dataset changes. We set attack methods (FGSM, Saliency Map) and noise level (0-19 with max pixel value 255) as independent variables and difference of prediction accuracy when noise was added as dependent variable in a simulation experiment. We have succeeded in extracting the threshold that separates non-adversarial and adversarial dataset. We detected the adversarial dataset using this threshold.

• Journal of the Ergonomics Society of Korea
• /
• v.31 no.6
• /
• pp.705-713
• /
• 2012

Objective: The aim of this study is to predict human visual target search using ACT-R cognitive architecture in real scene images. Background: Human uses both the method of bottom-up and top-down process at the same time using characteristics of image itself and knowledge about images. Modeling of human visual search also needs to include both processes. Method: In this study, visual target object search performance in real scene images was analyzed comparing experimental data and result of ACT-R model. 10 students participated in this experiment and the model was simulated ten times. This experiment was conducted in two conditions, indoor images and outdoor images. The ACT-R model considering the first saccade region through calculating the saliency map and spatial layout was established. Proposed model in this study used the guide of visual search and adopted visual search strategies according to the guide. Results: In the analysis results, no significant difference on performance time between model prediction and empirical data was found. Conclusion: The proposed ACT-R model is able to predict the human visual search process in real scene images using salience map and spatial layout. Application: This study is useful in conducting model-based evaluation in visual search, particularly in real images. Also, this study is able to adopt in diverse image processing program such as helper of the visually impaired.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.10 no.2
• /
• pp.769-786
• /
• 2016

Object recognition and object location have always drawn much interest. Also, recently various computational models have been designed. One of the big issues in this domain is the lack of an appropriate model for extracting important part of the picture and estimating the object place in the same environments that caused low accuracy. To solve this problem, a new Salient Based Bag of Visual Word (SBBoVW) model for object recognition and object location estimation is presented. Contributions lied in the present study are two-fold. One is to introduce a new approach, which is a Salient Based Bag of Visual Word model (SBBoVW) to recognize difficult objects that have had low accuracy in previous methods. This method integrates SIFT features of the original and salient parts of pictures and fuses them together to generate better codebooks using bag of visual word method. The second contribution is to introduce a new algorithm for finding object place based on the salient map automatically. The performance evaluation on several data sets proves that the new approach outperforms other state-of-the-arts.

• IEIE Transactions on Smart Processing and Computing
• /
• v.6 no.4
• /
• pp.229-236
• /
• 2017

This paper presents an efficient object-based color image-retrieval algorithm that is suitable for the classification and retrieval of images from small to mid-scale datasets, such as images in PCs, tablets, phones, and cameras. The proposed method first finds salient regions by using regional feature vectors, and also finds several dominant colors in each region. Then, each salient region is partitioned into small sub-blocks, which are assigned 1 or 0 with respect to the number of pixels corresponding to a dominant color in the sub-block. This gives a binary map for the dominant color, and this process is repeated for the predefined number of dominant colors. Finally, we have several binary maps, each of which corresponds to a dominant color in a salient region. Hence, the binary maps represent the spatial distribution of the dominant colors in the salient region, and the union (OR operation) of the maps can describe the approximate shapes of salient objects. Also proposed in this paper is a matching method that uses these binary maps and which needs very few computations, because most operations are binary. Experiments on widely used color image databases show that the proposed method performs better than state-of-the-art and previous color-based methods.

• Smart Media Journal
• /
• v.8 no.2
• /
• pp.39-45
• /
• 2019

As deep learning has become popular, researches which can help explaining the prediction results also become important. Superpixel based multi-scale combining technique, which provides the advantage of visual pleasing by maintaining the shape of the object, has been recently proposed. Based on the principle of prediction difference, this technique computes the saliency map from the difference between the predicted result excluding the superpixel and the original predicted result. In this paper, we propose a new technique of both excluding and including super pixels. Experimental results show 3.3% improvement in IoU evaluation.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.2
• /
• pp.771-789
• /
• 2019

Text detection has been a popular research topic in the field of computer vision. It is difficult for prevalent text detection algorithms to avoid the dependence on datasets. To overcome this problem, we proposed a novel unsupervised text detection algorithm inspired by bootstrap learning. Firstly, the text candidate in a novel form of superpixel is proposed to improve the text recall rate by image segmentation. Secondly, we propose a unique text sample selection model (TSSM) to extract text samples from the current image and eliminate database dependency. Specifically, to improve the precision of samples, we combine maximally stable extremal regions (MSERs) and the saliency map to generate sample reference maps with a double threshold scheme. Finally, a multiple kernel boosting method is developed to generate a strong text classifier by combining multiple single kernel SVMs based on the samples selected from TSSM. Experimental results on standard datasets demonstrate that our text detection method is robust to complex backgrounds and multilingual text and shows stable performance on different standard datasets.