We propose a hysteresis neural network system solving NP-Hard optimization problems, the N-Queens Problem. The continuous system with binary outputs searches a solution of the problem without energy function. The output vector corresponds to a complete solution when the output vector becomes stable. That is, this system does never become stable without satisfying the constraints of the problem. Though it is very hard to remove limit cycle completely from this system, we can propose a new method to reduce the possibility of limit cycle by controlling time constants.

Recently enhancing the visual experience of the user has been a new trend for TV displays. This trend comes from the fact that changes of ambient illuminations while viewing a Liquid Crystal Display (LCD) significantly affect human impressions. However, psychological effects caused by the combination of displayed video image and ambient illuminations have not been investigated. In the present research, we clarify the relationship between ambient illuminations and psychological effects while viewing video image displayed on the LCD by using a questionnaire based semantic differential (SD) method and a factor analysis method. Six kinds of video images were displayed under different colors and layouts of illumination conditions and rated by 15 observers. According to the analysis, it became clear that the illumination control around the LCD with displayed video image, the feeling of 'activity' and 'evaluating' were rated higher than the feeling of fluorescent ceiling condition. In particular, simultaneous illumination control around the display and the ceiling enhanced the feeling of 'activity,' and 'evaluating' with keeping 'comfort.' Moreover, the feeling of 'activity' under the illumination control around the LCD and the ceiling condition while viewing music video image was rated clearly higher than that with natural scene video image.

In medical fields, detecting traumatic bleedings has always been a difficult task due to the small size, low contrast of targets and large number of images. In this work we propose an automatic traumatic bleeding detection approach from contrast enhanced CT images via deep CNN networks, containing segmentation process and classification process. CT values of DICOM images are extracted and processed via three different window settings first. Small 3D patches are cropped from processed images and segmented by a 3D CNN network. Then segmentation results are converted to point cloud data format and classified by a classifier. The proposed pre-processing approach makes the segmentation network be able to detect small and low contrast targets and achieve a high sensitivity. The additional classification network solves the boundary problem and short-sighted problem generated during the segmentation process to further decrease false positives. The proposed approach is tested with 3 CT cases containing 37 bleeding regions. As a result, a total of 34 bleeding regions are correctly detected, the sensitivity reaches 91.89%. The average false positive number of test cases is 1678. 46.1% of false positive predictions are decreased after being classified. The proposed method is proved to be able to achieve a high sensitivity and be a reference of medical doctors.

In this paper, we propose a Pyramid Predictive Attention Network (PPAN) for medical image segmentation. In the medical field, the size of dataset generally restricts the performance of deep CNN and deploying the trained network with gross parameters into the terminal device with limited memory is an expectation. Our team aims to the future home medical diagnosis and search for lightweight medical image segmentation network. Therefore, we designed PPAN mainly made of Xception blocks which are modified from DeepLab v3+ and consist of separable depthwise convolutions to speed up the computation and reduce the parameters. Meanwhile, by utilizing pyramid predictions from each dimension stage will guide the network more accessible to optimize the training process towards the final segmentation target without degrading the performance. IoU metric is used for the evaluation on the test dataset. We compared our designed network performance with the current state of the art segmentation networks on our RGB tongue dataset which was captured by the developed TIAS system for tongue diagnosis. Our designed network reduced 80 percentage parameters compared to the most widely used U-Net in medical image segmentation and achieved similar or better performance. Any terminal with limited storage which is needed a segment of RGB image can refer to our designed PPAN.

Image quality of halftone print is significantly influenced by optical characteristics of paper. Light scattering in paper produces optical dot gain, which has a significant influence on the tone and color reproductions of halftone print. The light scattering can be quantified by the Modulation Transfer Function (MTF) of paper. Several methods have been proposed to measure the MTF of paper. However, these methods have problems in efficiency or accuracy in the measurement. In this article, a new method is proposed to measure the MTF of paper efficiently and accurately, and the dot gain effect on halftone print is analyzed. The MTF is calculated from the ratio in spatial frequency domain between the responses of incident pencil light to paper and the perfect specular reflector. Since the spatial frequency characteristic of input pencil light can be obtained from the response of perfect specular reflector, it does not need to produce the input illuminant having "ideal" impulse characteristic. Our method is experimentally efficient since only two images need to be measured. Besides it can measure accurately since the data can be approximated by the conventional MTF model. Next, we predict the reflectance distribution of halftone print using the measured MTF in microscopy in order to analyze the dot gain effect since it can clearly be observed in halftone micro-structure. Finally, a simulation is carried out to remove the light scattering effect from the predicted image. Since the simulated image is not affected by the optical dot gain, it can be applied to analyze the real dot coverage.

Due to the development and popularization of high-definition televisions, digital video cameras, Blu-ray discs, digital broadcasting, IP television and so on, it plays an important role to identify and quantify video quality degradations. In this paper, we propose SV-CIELAB which is an objective video quality assessment (VQA) method using a spatio-velocity contrast sensitivity function (SV-CSF). In SV-CIELAB, motion information in videos is effectively utilized for filtering unnecessary information in the spatial frequency domain. As the filter to apply videos, we used the SV-CSF. It is a modulation transfer function of the human visual system, and consists of the relationship among contrast sensitivities, spatial frequencies and velocities of perceived stimuli. In the filtering process, the SV-CSF cannot be directly applied in the spatial frequency domain because spatial coordinate information is required when using velocity information. For filtering by the SV-CSF, we obtain video frames separated in spatial frequency domain. By using velocity information, the separated frames with limited spatial frequencies are weighted by contrast sensitivities in the SV-CSF model. In SV-CIELAB, the criteria are obtained by calculating image differences between filtered original and distorted videos. For the validation of SV-CIELAB, subjective evaluation experiments were conducted. The subjective experimental results were compared with SV-CIELAB and the conventional VQA methods such as CIELAB color difference, Spatial-CIELAB, signal to noise ratio and so on. From the experimental results, it was shown that SV-CIELAB is a more efficient VQA method than the conventional methods.

Despite their usability advantages over traditional authentication systems, biometrics-based authentication systems suffer from inherent privacy violation and non-revocability issues. In order to address these issues, the concept of cancelable biometrics was introduced as a means of generating multiple, revocable, and noninvertible identities from true biometric templates. Apart from BioHashing, which is a two-factor cancelable biometrics technique based on mixing a set of tokenized user-specific random numbers with biometric features, cancelable biometrics techniques usually cannot preserve the recognition accuracy achieved using the unprotected biometric systems. However, as the employed token can be lost, shared, or stolen, BioHashing suffers from the same issues associated with token-based authentication systems. In this paper, a reliable tokenless cancelable biometrics scheme, referred to as BioEncoding, for protecting IrisCodes is presented. Unlike BioHashing, BioEncoding can be used as a one-factor authentication scheme that relies only on sole IrisCodes. A unique noninvertible compact bit-string, referred to as BioCode, is randomly derived from a true IrisCode. Rather than the true IrisCode, the derived BioCode can be used efficiently to verify the user identity without degrading the recognition accuracy obtained using original IrisCodes. Additionally, BioEncoding satisfies all the requirements of the cancelable biometrics construct. The performance of BioEncoding is compared with the performance of BioHashing in the stolen-token scenario and the experimental results show the superiority of the proposed method over BioHashing-based techniques.

The block based motion estimation technique is adopted by various video coding standards to reduce the temporal redundancy in video sequences. The core of that technique is the search algorithm implemented to find the location of the best matched block. Indeed, the full search algorithm is the most straightforward and optimal but computationally demanding search algorithm. Consequently, many fast and suboptimal search algorithms have been proposed. Reduction of the number of location being searched is the approach used to decrease the computational load of full search. In this paper, hybridization between an adaptive search algorithm and the full search algorithm is proposed. The adaptive search algorithm benefits from the correlation within spatial and temporal adjacent blocks. At the same time, a feature domain based matching criteria is used to reduce the complexity resulting from applying the pixel based conventional criteria. It is shown that the proposed algorithm produces good quality performance and requires less computational time compared with popular block matching algorithms.

In this paper, we propose a new projector-based display which can perform the color simulator for print industry. The proposed color simulator can change the color of print by projecting the image onto the print. A color of print can be matched to the desired color by projecting the image which is calculated to minimize the color difference between the colors of target print and current print. This current print is measured by digital camera or digital scanner. Ideally, spectral camera or scanner is expected to be used for accurate color simulation on the current print, but it costs a lot for practical application. Therefore, in this paper, we compared two methods for color matching, one is the tristimulus-based method with XYZ tristimulus values and the other is the spectral-based method with spectral values. As the result of computer simulation, the average color difference ΔE *94 was 0.27 by the spectral-based method between the reflected radiance from the color of target print and the color of current print with projector, and the average color difference ΔE *94 was 2.09 by the tristimulus-based method. The efficiency of the proposed system is verified by the subjective evaluation between the target and current print with appropriate image projection.

Proving the security of cancelable biometrics and other template protection techniques is a key prerequisite for the widespread deployment of biometric technologies. BioEncoding is a cancelable biometrics scheme that has been proposed recently to protect biometric templates represented as binary strings like iris codes. Unlike other template protection schemes, BioEncoding does not require user-specific keys or tokens. Moreover, it satisfies the requirements of untraceable biometrics without sacrificing the matching accuracy. However, the security of BioEncoding against smart attacks, such as correlation and optimization-based attacks, has to be proved before recommending it for practical deployment. In this paper, the security of BioEncopding, in terms of both non-invertibility and privacy protection, is analyzed. First, resistance of protected templates generated using BioEncoding against brute-force search attacks is revisited rigorously. Then, vulnerabilities of BioEncoding with respect to correlation attacks and optimization based attacks are identified and explained. Furthermore, an important modification to the BioEncoding algorithm is proposed to enhance its security against correlation attacks. The effect of integrating this modification into BioEncoding is validated and its impact on the matching accuracy is investigated empirically using CASIA-IrisV3-Interval dataset. Experimental results confirm the efficacy of the proposed modification and show that it has no negative impact on the matching accuracy.

The purpose of this research is to propose an effective color metric which can predict the perceptual image quality for Retinex method. In this paper, we first give a brief introduction of three kinds of typical single Retinex methods to improve the color reproduction. And then, we state the process for obtaining the observer rating value from the subjective evaluation experiment performed under the sRGB illumination condition. Next, we introduce the S-CIELAB metric and propose a new metric on the basis of S-CIELAB metric that considers the gazing information. The average S-CIELAB color differences with and without the consideration of gazing information were calculated as the objective image quality measures. The correlations between the observer rating values and the objective image quality measures were calculated. The result shows that all of the average S-CIELAB color differences based on the gazing information are better correlated to the observer rating value than the average S-CIELAB color difference over the whole area. The average S-CIELAB color difference weighted by the gazing frequency over the gazing area shows the strong correlation with the observer rating value.

The aim of this work is to develop an efficient medical image segmentation technique by fitting a nonlinear shape model with pre-segmented images. In this technique, the kernel principle component analysis (KPCA) is used to capture the shape variations and to build the nonlinear shape model. The pre-segmentation is carried out by classifying the image pixels according to the high level texture features extracted using the over-complete wavelet packet decomposition. Additionally, the model fitting is completed using the particle swarm optimization technique (PSO) to adapt the model parameters. The proposed technique is fully automated, is talented to deal with complex shape variations, can efficiently optimize the model to fit the new cases, and is robust to noise and occlusion. In this paper, we demonstrate the proposed technique by implementing it to the liver segmentation from computed tomography (CT) scans and the obtained results are very hopeful.

We investigated the relationship between ambient illumination and psychological effect by applying a modified color harmony model. We verified the proposed model by analyzing correlation between psychological value and modified color harmony score. Experimental results showed the possibility to obtain the best color for illumination using this model.

We propose hysteresis neural network solving combinatorial optimization problems, Box Puzzling Problem. Hysteresis neural network searches solutions of the problem with nonlinear dynamics. The output vector becomes stable only when it corresponds with a solution. This system does never become stable without satisfying constraints of the problem. After estimating hardware calculating time, we obtain that numerical calculating time increases extremely comparing with hardware time as problem's scale increases. However the system has possibility of limit cycle. Though it is very hard to remove limit cycle completely, we propose some methods to remove this phenomenon.