Hyunjun Eun

No-Reference Image Quality Assessment Based on Singular Value Decomposition Without Learning

Recently no-reference image quality assessment NR-IQA methods take advantages of machine learning techniques. However, machine learning approaches need a number of human scored images and cause database dependency. In this paper, we propose a simple NR-IQA method that can estimate quality of distorted images without learning, producing comparable performance to learning based approaches. We employ singular value decomposition SVD since we have observed that singular values are commonly affected by various distortions. In detail, a decreasing rate of singular values is highly correlated to a degree of distortions regardless of their type. From the observation, our approach utilizes the decreasing rate of singular values to model a simple and reliable NR-IQA method. Experimental results show that the proposed method has reasonably high correlation to human scores. And the proposed method can secure simplicity and database independence.

Superpixel-Guided Adaptive Image Smoothing

In edge-preserving image smoothing, edge blurriness and structural edge attenuation have been common problems. L0 smoothing successfully solves these two problems by adopting L0 norm of gradients. However, a weak structural edge diminishing problem still exists because L0 penalty first removes small nonzero gradients. In order to address this problem, we propose superpixel-guided adaptive image smoothing by introducing an adaptive parameter into L0 smoothing framework. The adaptive smoothing parameter is efficiently computed in a cascade manner. In the first stage, we allocate smoothing parameters to the pixels consisting of details. More importantly, we then exploit similarities between a pixel and its surrounding superpixels for assigning smoothing parameters to the rest pixels. Experimental results demonstrate that our proposed method efficiently preserves structural edges regardless of their scales compared to previous methods.

Boosting Proximal Dental Caries Detection via Combination of Variational Methods and Convolutional Neural Network

Proximal dental caries are diagnosed using dental X-ray images. Unfortunately, the diagnosis of proximal dental caries is often stifled due to the poor quality of dental X-ray images. Therefore, we propose an automatic detection system to detect proximal dental caries in periapical images for the first time. The system comprises four modules: horizontal alignment of pictured teeth, probability map generation, crown extraction, and refinement. We first align the pictured teeth horizontally as a pre-process to minimize performance degradation due to rotation. Next, a fully convolutional network are used to produce a caries probability map while crown regions are extracted based on optimization schemes and an edge-based level set method. In the refinement module, the caries probability map is refined by the distance probability modeled by crown regions since caries are located near tooth surfaces. Also we adopt non-maximum suppression to improve the detection performance. Experiments on various periapical images reveal that the proposed system using a convolutional neural network (CNN) and crown extraction is superior to the system using a naïve CNN.

Saliency refinement: Towards a uniformly highlighted salient object

Abstract Humans have a natural tendency to view a visually attractive (i.e., salient) object in its entirety. However, previous methods for salient object detection only highlight some parts of the salient object. This problem severely limits the adoption of such technologies to various computer vision and pattern recognition applications. To address the problem, in this paper, we present a novel framework to improve a saliency map obtained from recent state-of-the-art salient object detection approaches. Based on the fact that the L 0 optimization can efficiently minimize variation between values, we integrate a background saliency and an initial saliency based on the nonlocal L 0 optimization. In our work, we first extract background samples to estimate the background saliency building upon the initial saliency and color information. We then integrate the background saliency into the initial saliency by solving an optimization problem. We formulate the optimization problem based on the nonlocal L 0 gradient to efficiently minimize the saliency variation in the salient object. To confirm the effectiveness of our proposed method, we apply the proposed framework to the saliency maps generated from state-of-the-art methods. Experimental results on benchmark datasets demonstrate that the proposed framework significantly improves the saliency maps. Furthermore, we compare the performance of two refinement frameworks and ours to prove the superiority of our work.

Single-view 2D CNNs with fully automatic non-nodule categorization for false positive reduction in pulmonary nodule detection

BACKGROUND AND OBJECTIVE In pulmonary nodule detection, the first stage, candidate detection, aims to detect suspicious pulmonary nodules. However, detected candidates include many false positives and thus in the following stage, false positive reduction, such false positives are reliably reduced. Note that this task is challenging due to 1) the imbalance between the numbers of nodules and non-nodules and 2) the intra-class diversity of non-nodules. Although techniques using 3D convolutional neural networks (CNNs) have shown promising performance, they suffer from high computational complexity which hinders constructing deep networks. To efficiently address these problems, we propose a novel framework using the ensemble of 2D CNNs using single views, which outperforms existing 3D CNN-based methods. METHODS Our ensemble of 2D CNNs utilizes single-view 2D patches to improve both computational and memory efficiency compared to previous techniques exploiting 3D CNNs. We first categorize non-nodules on the basis of features encoded by an autoencoder. Then, all 2D CNNs are trained by using the same nodule samples, but with different types of non-nodules. By extending the learning capability, this training scheme resolves difficulties of extracting representative features from non-nodules with large appearance variations. Note that, instead of manual categorization requiring the heavy workload of radiologists, we propose to automatically categorize non-nodules based on the autoencoder and k-means clustering. RESULTS We performed extensive experiments to validate the effectiveness of our framework based on the database of the lung nodule analysis 2016 challenge. The superiority of our framework is demonstrated through comparing the performance of five frameworks trained with differently constructed training sets. Our proposed framework achieved state-of-the-art performance (0.922 of the competition performance metric score) with low computational demands (789K of parameters and 1024M of floating point operations per second). CONCLUSION We presented a novel false positive reduction framework, the ensemble of single-view 2D CNNs with fully automatic non-nodule categorization, for pulmonary nodule detection. Unlike previous 3D CNN-based frameworks, we utilized 2D CNNs using 2D single views to improve computational efficiency. Also, our training scheme using categorized non-nodules, extends the learning capability of representative features of different non-nodules. Our framework achieved state-of-the-art performance with low computational complexity.

Oriented tooth localization for periapical dental X-ray images via convolutional neural network

In studies using dental X-ray images, tooth localization is essential to produce accurate results. In this paper, we propose a tooth localization method that tightly localize diverse teeth in periapical dental X-ray images. Oriented tooth proposals are generated by using teeth separation lines (TSLs) and a tooth top line, which are reliable and tight to teeth. To classify each tooth proposal into either a tooth or a non-tooth, we utilize a convolutional neural network (CNN). Our CNN model is trained with three classes, i.e., one negative and two positives, for better classification. In addition, we propose scale based non-maximum suppression by integrating scale confidence with non-maximum suppression to efficiently eliminate multiple tooth localizations.

Salient object detection using HOS based L0 smoothing and shape-aware region merging

Recent salient object detection algorithms often involve a segmentation step to produce saliency maps preserving boundaries. However, over-segmented results that many segmentation methods produce confuse to describe object boundaries. In this paper, we present a novel salient object detection algorithm which produces reliable salient object candidates. First, the input image is processed by Higher Order Statistics (HOS) based L0 smoothing to highlight strong edges and reduce texture. We then apply image segmentation to the HOS L0 smoothed image to produce improved results in which the number of over-segmented regions is greatly reduced. Second, we propose shape-aware region merging with a novel region scale measure. Finally, a saliency map from the merging result is generated by taking two simple saliency cues. Extensive experiments on a challenging saliency dataset indicate that our algorithm has comparable performance against state-of-the-arts.