Variational Deep Image Restoration.

This paper presents a new variational inference framework for image restoration and a convolutional neural network (CNN) structure that can solve the restoration problems described by the proposed framework. Earlier CNN-based image restoration methods primarily focused on network architecture design or training strategy with non-blind scenarios where the degradation models are known or assumed. For a step closer to real-world applications, CNNs are also blindly trained with the whole dataset, including diverse degradations. However, the conditional distribution of a high-quality image given a diversely degraded one is too complicated to be learned by a single CNN. Therefore, there have also been some methods that provide additional prior information to train a CNN. Unlike previous approaches, we focus more on the objective of restoration based on the Bayesian perspective and how to reformulate the objective. Specifically, our method relaxes the original posterior inference problem to better manageable sub-problems and thus behaves like a divide-and-conquer scheme. As a result, the proposed framework boosts the performance of several restoration problems compared to the previous ones. Specifically, our method delivers state-of-the-art performance on Gaussian denoising, real-world noise reduction, blind image super-resolution, and JPEG compression artifacts reduction. Our code and more details are available on our project page, https://github.com/JWSoh/VDIR.

Inverse-Based Approach to Explaining and Visualizing Convolutional Neural Networks.

This article presents a new method for understanding and visualizing convolutional neural networks (CNNs). Most existing approaches to this problem focus on a global score and evaluate the pixelwise contribution of inputs to the score. The analysis of CNNs for multilabeled outputs or regression has not yet been considered in the literature, despite their success on image classification tasks with well-defined global scores. To address this problem, we propose a new inverse-based approach that computes the inverse of a feedforward pass to identify activations of interest in lower layers. We developed a layerwise inverse procedure based on two observations: 1) inverse results should have consistent internal activations to the original forward pass and 2) a small amount of activation in inverse results is desirable for human interpretability. Experimental results show that the proposed method allows us to analyze CNNs for classification and regression in the same framework. We demonstrated that our method successfully finds attributions in the inputs for image classification with comparable performance to state-of-the-art methods. To visualize the tradeoff between various methods, we developed a novel plot that shows the tradeoff between the amount of activations and the rate of class reidentification. In the case of regression, our method showed that conventional CNNs for single image super-resolution overlook a portion of frequency bands that may result in performance degradation.

Co-Salient Object Detection Based on Deep Saliency Networks and Seed Propagation Over an Integrated Graph.

This paper presents a co-salient object detection method to find common salient regions in a set of images. We utilize deep saliency networks to transfer co-saliency prior knowledge and better capture high-level semantic information. The resulting initial co-saliency maps are enhanced by seed propagation steps over an integrated graph. The deep saliency networks are trained in a supervised manner to avoid weakly supervised online learning and exploit them not only to extract high-level features but also to produce both intra- and inter-image saliency maps. Through a refinement step, the initial co-saliency maps can uniformly highlight co-salient regions and locate accurate object boundaries. To handle input image groups inconsistent in size, we propose to pool multi-regional descriptors including both within-segment and within-group information. In addition, the integrated multilayer graph is constructed to find the regions that the previous steps may not detect by seed propagation with low-level descriptors. In this paper, we utilize the useful complementary components of high- and low-level information and several learning-based steps. Our experiments have demonstrated that the proposed approach outperforms comparable co-saliency detection methods on widely used public databases and can also be directly applied to co-segmentation tasks.

Does Sacral Slanting Affect Distal Adding-on in Lenke Type 1A Adolescent Idiopathic Scoliosis?

STUDY DESIGN: A retrospective clinical study. OBJECTIVE: The purpose of this study was to identify risk factors for postoperative distal adding-on in Lenke 1A adolescent idiopathic scoliosis. SUMMARY OF BACKGROUND DATA: Distal adding-on is a postoperative complication associated with the Lenke type 1A curve. Although various factors are known to cause postoperative adding-on, no study has reported a correlation between sacral slanting and adding-on. METHODS: A total of 126 consecutive patients who underwent posterior correction and fusion surgery for Lenke type 1A adolescent idiopathic scoliosis were included in this study. Curve type was further categorized into L4-left (L4-L) or L4-right (L4-R), based on the direction of the L4 vertebral tilt. Several clinical and radiological parameters including sacral slanting were investigated to identify risk factors associated with postoperative distal adding-on. RESULTS: A total of 36 patients (28.6%) exhibited sacral slanting. Nineteen out of 20 L4-L type patients had left-sided sacral slanting, whereas 12 out of 16 L4-R type patients had right-sided sacral slanting. The group with adding-on (n = 13) demonstrated a significantly lower age than the group without adding-on (n = 113). Preoperative lumbar Cobb angle (P = 0.022) was determined to be an independent factor for adding-on according to logistic regression analysis. In the L4-R type, the last touching vertebra (LTV) level and the gap difference in levels between lowest instrumented vertebra and LTV (lowest instrumented vertebra-LTV) comprised significant variables. CONCLUSION: Sacral slanting typically occurs to the left in the L4-L type and to the right in the L4-R type. The size of the preoperative lumbar curve was found to be an independent risk factor that caused adding-on in patients with Lenke type 1A scoliosis. In the L4-R type, right-sided sacral slanting tended to lower the LTV. Therefore, the fusion level might be shorter to save the motion segments resulting in distal adding-on. LEVEL OF EVIDENCE: 4.

Referral patterns and patient characteristics at the first visit to a scoliosis center: a 2-year experience in South Korea without a school scoliosis-screening program.

OBJECTIVE The need for scoliosis screening remains controversial. Nationwide school screening for scoliosis has not been performed in South Korea, and there are few studies on the referral patterns of patients suspected of having scoliosis. This study aimed to examine the referral patterns to the largest scoliosis center in South Korea in the absence of a school screening program and to analyze the factors that influence the appropriateness of referral. METHODS The medical records of patients who visited a single scoliosis center for a spinal deformity evaluation were reviewed. Among 1895 new patients who visited this scoliosis center between April 2014 and March 2016, 1211 with presumed adolescent idiopathic scoliosis were included in the study. Patients were classified into 4 groups according to the referral method: non-health care provider, primary physician, hospital specialist, or school screening program. The appropriateness of referral was labeled as inappropriate, late, or appropriate. In total, 213 of 1211 patients were excluded because they had received treatment at another medical facility; 998 patients were evaluated to determine the appropriateness of referral. RESULTS Of the 998 referrals of new patients with presumed adolescent idiopathic scoliosis, 162 (16.2%) were classified as an inappropriate referral, 272 (27.3%) were classified as a late referral, and 564 (56.5%) were classified as an appropriate referral. Age, sex, Cobb angle of the major curve, and skeletal maturity were identified as statistically significant factors that correlated with the appropriateness of referral. The referral method did not correlate with the appropriateness of referral. CONCLUSIONS Under the current health care system in South Korea, a substantial number of patients with presumed adolescent idiopathic scoliosis are referred either late or inappropriately to a tertiary medical center. Although patients referred by school screening programs had a significantly lower late referral rate and higher appropriate referral rate than the other 3 groups, the referral method was not a significant factor in terms of the appropriateness of referral.

Optimal Power Allocation of Relay Sensor Node Capable of Energy Harvesting in Cooperative Cognitive Radio Network.

A cooperative cognitive radio scheme exploiting primary signals for energy harvesting is proposed. The relay sensor node denoted as the secondary transmitter (ST) harvests energy from the primary signal transmitted from the primary transmitter, and then uses it to transmit power superposed codes of the secrecy signal of the secondary network (SN) and of the primary signal of the primary network (PN). The harvested energy is split into two parts according to a power splitting ratio, one for decoding the primary signal and the other for charging the battery. In power superposition coding, the amount of fractional power allocated to the primary signal is determined by another power allocation parameter (e.g., the power sharing coefficient). Our main concern is to investigate the impact of the two power parameters on the performances of the PN and the SN. Analytical or mathematical expressions of the outage probabilities of the PN and the SN are derived in terms of the power parameters, location of the ST, channel gain, and other system related parameters. A jointly optimal power splitting ratio and power sharing coefficient for achieving target outage probabilities of the PN and the SN, are found using these expressions and validated by simulations.

Semi-Circumferential Decompression: Microsurgical Total en-bloc Ligamentum Flavectomy to Treat Lumbar Spinal Stenosis with Grade I Degenerative Spondylolisthesis.

BACKGROUND: To describe and assess clinical outcomes of the semi-circumferential decompression technique for microsurgical en-bloc total ligamentum flavectomy with preservation of the facet joint to treat the patients who have a lumbar spinal stenosis with degenerative spondylolisthesis. METHODS: We retrospectively analyzed the clinical and radiologic outcomes of 19 patients who have a spinal stenosis with Meyerding grade I degenerative spondylolisthesis. They were treated using the "semi-circumferential decompression" method. We evaluated improvements in back and radiating pain using a visual analogue scale (VAS) and the Oswestry Disability Index (ODI). We also evaluated occurrence of spinal instability on radiological exam using percentage slip and slip angle. RESULTS: The mean VAS score for back pain decreased significantly from 6.3 to 4.3, although some patients had residual back pain. The mean VAS for radiating pain decreased significantly from 8.3 to 2.5. The ODI score improved significantly from 25.3 preoperatively to 10.8 postoperatively. No significant change in percentage slip was observed (10% preoperatively vs. 12.2% at the last follow-up). The dynamic percentage slip (gap in percentage slip between flexion and extension X-ray exams) did not change significantly (5.2% vs. 5.8%). Slip angle and dynamic slip angle did not change (3.2° and 8.2° vs. 3.6° and 9.2°, respectively). CONCLUSIONS: The results suggested that semi-circumferential decompression is a clinically recommendable procedure that can improve pain. This procedure does not cause spinal instability when treating patients who have a spinal stenosis with degenerative spondylolisthesis.

Hierarchical prediction and context adaptive coding for lossless color image compression.

This paper presents a new lossless color image compression algorithm, based on the hierarchical prediction and context-adaptive arithmetic coding. For the lossless compression of an RGB image, it is first decorrelated by a reversible color transform and then Y component is encoded by a conventional lossless grayscale image compression method. For encoding the chrominance images, we develop a hierarchical scheme that enables the use of upper, left, and lower pixels for the pixel prediction, whereas the conventional raster scan prediction methods use upper and left pixels. An appropriate context model for the prediction error is also defined and the arithmetic coding is applied to the error signal corresponding to each context. For several sets of images, it is shown that the proposed method further reduces the bit rates compared with JPEG2000 and JPEG-XR.

Text-line extraction in handwritten Chinese documents based on an energy minimization framework.

Text-line extraction in unconstrained handwritten documents remains a challenging problem due to nonuniform character scale, spatially varying text orientation, and the interference between text lines. In order to address these problems, we propose a new cost function that considers the interactions between text lines and the curvilinearity of each text line. Precisely, we achieve this goal by introducing normalized measures for them, which are based on an estimated line spacing. We also present an optimization method that exploits the properties of our cost function. Experimental results on a database consisting of 853 handwritten Chinese document images have shown that our method achieves a detection rate of 99.52% and an error rate of 0.32%, which outperforms conventional methods.

Design of interchannel MRF model for probabilistic multichannel image processing.

In this paper, we present a novel framework that exploits an informative reference channel in the processing of another channel. We formulate the problem as a maximum a posteriori estimation problem considering a reference channel and develop a probabilistic model encoding the interchannel correlations based on Markov random fields. Interestingly, the proposed formulation results in an image-specific and region-specific linear filter for each site. The strength of filter response can also be controlled in order to transfer the structural information of a channel to the others. Experimental results on satellite image fusion and chrominance image interpolation with denoising show that our method provides improved subjective and objective performance compared with conventional approaches.

Composition of a dewarped and enhanced document image from two view images.

In this paper, we propose an algorithm to compose a geometrically dewarped and visually enhanced image from two document images taken by a digital camera at different angles. Unlike the conventional works that require special equipment or assumptions on the contents of books or complicated image acquisition steps, we estimate the unfolded book or document surface from the corresponding points between two images. For this purpose, the surface and camera matrices are estimated using structure reconstruction, 3-D projection analysis, and random sample consensus-based curve fitting with the cylindrical surface model. Because we do not need any assumption on the contents of books, the proposed method can be applied not only to optical character recognition (OCR), but also to the high-quality digitization of pictures in documents. In addition to the dewarping for a structurally better image, image mosaic is also performed for further improving the visual quality. By finding better parts of images (with less out of focus blur and/or without specular reflections) from either of views, we compose a better image by stitching and blending them. These processes are formulated as energy minimization problems that can be solved using a graph cut method. Experiments on many kinds of book or document images show that the proposed algorithm robustly works and yields visually pleasing results. Also, the OCR rate of the resulting image is comparable to that of document images from a flatbed scanner.