Noriko Yata

Automatic construction of image transformation algorithms using feature based genetic image network

Image processing and recognition technologies are becoming increasingly important. Automatic construction methods for image transformation algorithms proposed to date approximate adequate image transformation from original images to their target images using a combination of several known image processing filters by evolutionary computation techniques. In this paper, we introduce the adaptive image processing filters that process according to the features of an input image. The processing of the adaptive filters is decided based on the local features of an input image. We implement them to feed-forward genetic image network (FFGIN) that is one of the automatic construction methods for image transformations. Then we apply our method to the problems of segmentation of organs and tissues in medical images. Experimental results show that our method constructs the effective segmentation algorithms that extract multiple regions respectively.

Evolving search spaces to emphasize the performance difference of real-coded crossovers using genetic programming

When we evaluate the search performance of an evolutionary computation (EC) technique, we usually apply it to typical benchmark functions and evaluate its performance in comparison to other techniques. In experiments on limited benchmark functions, it can be difficult to understand the features of each technique. In this paper, the search spaces that emphasize the performance difference of EC techniques are evolved by Cartesian genetic programming. We focus on a real-coded genetic algorithm, which is a type of genetic algorithm that has a real-valued vector as a chromosome. In particular, we generate search spaces using the performance difference of real-coded crossovers. In the experiments, we evolve the search spaces using the combination of three types of real-coded crossovers. As a result of our experiments, the search spaces that exhibit the largest performance difference of two crossovers are generated for all the combinations.

Deep learning analyzes Helicobacter pylori infection by upper gastrointestinal endoscopy images

Background and study aims  Helicobacter pylori (HP)-associated chronic gastritis can cause mucosal atrophy and intestinal metaplasia, both of which increase the risk of gastric cancer. The accurate diagnosis of HP infection during routine medical checks is important. We aimed to develop a convolutional neural network (CNN), which is a machine-learning algorithm similar to deep learning, capable of recognizing specific features of gastric endoscopy images. The goal behind developing such a system was to detect HP infection early, thus preventing gastric cancer. Patients and methods  For the development of the CNN, we used 179 upper gastrointestinal endoscopy images obtained from 139 patients (65 were HP-positive: ≥ 10 U/mL and 74 were HP-negative: < 3 U/mL on HP IgG antibody assessment). Of the 179 images, 149 were used as training images, and the remaining 30 (15 from HP-negative patients and 15 from HP-positive patients) were set aside to be used as test images. The 149 training images were subjected to data augmentation, which yielded 596 images. We used the CNN to create a learning tool that would recognize HP infection and assessed the decision accuracy of the CNN with the 30 test images by calculating the sensitivity, specificity, and area under the receiver operating characteristic (ROC) curve (AUC). Results  The sensitivity and specificity of the CNN for the detection of HP infection were 86.7 % and 86.7 %, respectively, and the AUC was 0.956. Conclusions  CNN-aided diagnosis of HP infection seems feasible and is expected to facilitate and improve diagnosis during health check-ups.

Diminishing Real Objects and Adding Virtual Objects Using a RGB-D Camera

AR technology is often used in applications that simulate an arrangement of furniture. These applications superimpose CG furnitures on any place. These applications cannot replace the CG furniture with a real furniture if there is a real furniture. To solve this problem, this paper proposes a method that can erase real object from the real environment (Diminished Reality) and add virtual object to the real environment (Augmented Reality) including the region of erased object using RGB-D camera.

Design of an AR marker for cylindrical surface

This paper proposes an augmented reality marker that can be robustly detected even on a cylindrical surface. The marker enables the surface normal estimation of a cylindrical object to realize the presentation of appropriate virtual information on the object. Conventional markers have difficulty detecting and obtaining accurate surface normal in the presence of occlusion or distortion of the marker in the image. Furthermore, it is difficult to identify a feature on a cylindrical object on which to position a marker. These problems are resolved by relying on the characteristic that a line parallel to the central axis of the cylinder maintains linearity. In addition, surface normal is calculated by estimating the objects shape by using transformation matrices.

Interactive Estimation of Light Source Position and Reflectance of Real Objects for Mixed-Reality Application

The seamless integration of real and virtual objects is required for mixed-reality applications. To achieve this goal, we should represent an effect of light reflection like shading, shadowing and inter- reflection between the real and virtual objects. To represent these effects, we have to estimate reflectance of the real objects. The reflectance can be estimated with color and geometry of the objects and light condition of the scene. To calculate at an interactive frame rates, the light sources are distributed on a surface of a dome above the scene. To estimate the reflectance more accurately, we have to calculate distance from the objects to the light source. Therefore, this paper proposes a method to estimate the distance from the objects to the light source and the reflectance of the objects at an interactive frame rates. In the proposed method, two cameras and a marker with a spherical mirror are used. We can use an RGB camera and an IR camera of Microsoft Kinect sensor as the cameras. In other words, by use of the proposed method, we can estimate the distance and reflectance by using the Kinect and the marker with the spherical mirror. In the method, intersection points of reflection vectors on the spherical mirror at each camera are evaluated and the point which has the maximum evaluation value is regarded as an estimation value of the light source position. With the proposed method, we can estimate the light source position and reflectance of the real objects at an interactive frame rates by use of the Kinect and the marker with the spherical mirror, so that we are able to apply the method to various mixed-reality applications.

Ensemble image classification method based on genetic image network

Automatic construction method for image classification algorithms have been required. Genetic Image Network for Image Classification (GIN-IC) is one of the methods that construct image classification algorithms automatically, and its effectiveness has already been proven. In our study, we try to improve the performance of GIN-IC with AdaBoost algorithm using GIN-IC as weak classifiers to complement with each other. We apply our proposed method to three types of image classification problems, and show the results in this paper. In our method, discrimination rates for training images and test images improved in the experiments compared with the previous method GIN-IC.

[POSTER] Consistency between Reflection on the Glass and Virtual Object in Augmented Reality

Augmented Reality (AR) technology is often used for arrangement simulation such as furniture before purchasing. This study focuses on the simulation of product layout in stores and exhibits in museums with show-window. Displaying the virtual information with correct position against real objects is important for realistic experience in AR applications. In this study, we keep the consistency between reflection on the glass and virtual object. Consequently, we extract the reflections on the glass using stereo images according to disparity, and represent it on the superimposed CG object in real-time. Experimental result shows the appearance of the CG object compared with that of the real object in showwindow.

Acquisition of image features for material perception from fine-tuned convolutional neural networks

Many studies of material property estimation and material recognition have been conducted. Previous approaches evaluate the validity or usefulness of hand-designed image features. Thus, we propose a method to directly and naturally acquire image features for material perception using convolutional neural networks. Using a fine-tuned network, we achieved approximately the same recognition accuracy as recent results on the Flickr Material Database. We found that neurons in deep layers are trained to respond to textures or materials.

[Poster] Representing degradation of real objects using augmented reality

Much research in augmented reality (AR) technology attempts to match the textures of virtual objects with real world. However, the textures of real objects must also be rendered consistent with virtual information. This paper proposes a method for representing the degradation of real objects in virtual time. Real-world depth information, used to build three-dimensional models of real objects, is captured by a RGB-D camera. The degradation of real objects is then represented by superimposing the degradation texture onto the real object.