Wook-Hyun Kim

Self-organization neural network for multiple texture image segmentation

Texture analysis is an important technique in many image processing areas, such as scene segmentation, object recognition, and shape and depth perception. But no efficient methods captures all aspects of the very diverse texture family including natural scenes. We propose a novel approach for efficient texture image analysis that use unsupervised learning schemes for the texture recognition task. The self-organization neural network for texture image identification is based on features that is extracted at angle and magnitude in the orientation-field that might be different from the sample textures. In order to show the performance of the proposed system, we have attempted to build various texture images. The experimental results show that the performance of the system is very successful.

Segmentation and recognition of traffic signs using shape information

This paper proposes a method for traffic sign recognition and segmentation using shape information of traffic sign. First, a connected component algorithm is used to segment candidate traffic sign regions from a binary image obtained based on the RGB color ratio of each pixel in image. Then actual traffic sign regions are segmented based on their X- and Y-axes symmetry. The recognition step utilizes shape information, including a moment, edge correlogram, and the number of times a concentric circular pattern from the region center intersects with the frequency information extracted by the wavelet transform. Finally, recognition is performed by measuring the similarity with templates in a database. Experimental results confirm the validity of the proposed method as regards geometric transformations and environmental factors.

Multi-level thresholding using entropy-based weighted FCM algorithm in color image

This paper proposes a multi-level thresholding method based on a weighted FCM(Fuzzy C-Means) algorithm in color image. FCM algorithm can determine a more optimal thresholding value than existing methods and be extended to multi-level thresholding, yet it is sensitive to noise, as it does not include spatial information. To solve this problem, a weight based on the entropy obtained from neighboring pixels is applied to FCM algorithm, and the optimal cluster number is determined using the within-class distance in the code image based on the clustered pixels for each color component. Experiments confirmed that the proposed method was more tolerant to noise and superior to existing methods.

Texture segmentation by unsupervised learning and histogram analysis using boundary tracing

Texture analysis is an important technique in many image processing areas, such as scene segmentation, object recognition, and shape&depth perception. However, most methods are restricted to issue of computational complexity and supervised problems. Accordingly, we propose a efficient method of segmenting texture that uses unsupervised learning schemes to discover a texture cluster without a pre-knowledge. This method applies 2D Gaussian filters to the clustered region iteratively, and the thresholding value for segmenting is automatically determined by analyzing histogram of the clustered inner-region. It can be acquired by the boundary tracing in the clustered region. In order to show the performance of the proposed method, we have attempted to build a various texture images, and the segmenting quality was measured according to the goodness based on the segmented shape of region. Our experimental results showed that the performance of the proposed method is very successful.

EWFCM algorithm and region-based multi-level thresholding

Multi-level thresholding is a method that is widely used in image segmentation. However, most of the existing methods are not suited to be directly used in applicable fields, and moreover they are not extended into a step of image segmentation. This paper proposes region-based multi-level thresholding as an image segmentation method. At first, we classify pixels of each color channel to two clusters by using EWFCM algorithm that is an improved FCM algorithm with spatial information between pixels. To obtain better segmentation results, a reduction of clusters is then performed by a region-based reclassification step based on a similarity between regions existing in a cluster and the other clusters. We finally perform a region merging by Bayesian algorithm based on Kullback-Leibler distance between a region and the neighboring regions as a post-processing method, as many regions still exist in image. Experiments show that region-based multi-level thresholding is superior to cluster-, pixel-based multi-level thresholding, and an existing method and much better segmentation results are obtained by the proposed post-processing method.

Segmentation and Recognition of Traffic Signs using Shape Information and Edge Image in Real Image

This study proposes a method for segmentation and recognition of traffic signs using shape information and edge image in real image. It first segments traffic sign candidate regions by connected component algorithm from binary images, obtained by utilizing the RGB color ratio of each pixel in the image, and then extracts actual traffic signs based on their symmetries on X- and Y-axes. Histogram equalization is performed for unsegmented candidate regions caused by low contrast in the image. In the recognition stage, it utilizes shape information including projection profiles on X- and Y-axes, moment, and the number of crossings and distance which concentric circular patterns and 8-directional rays from region center intersects with edges of traffic signs. It finally performs recognition by measuring similarity with the templates in the database. It will be shown from several experimental results that the system is robust to environmental factors, such as light and weather condition.

Multiple Pedestrians Detection and Tracking using Color Information from a Moving Camera

This paper presents a new method for the detection of multiple pedestrians and tracking of a specific pedestrian using color information from a moving camera. We first extract motion vector on the input image using BMA. Next, a difference image is calculated on the basis of the motion vector. The difference image is converted to a binary image. The binary image has an unnecessary noise. So, it is removed by means of the proposed noise deletion method. Then, we detect pedestrians through the projection algorithm. But, if pedestrians are very adjacent to each other, we separate them using RGB color information. And we track a specific pedestrian using RGB color information in center region of it. The experimental results on our test sequences demonstrated the high efficiency of our approach as it had shown detection success ratio of 97％ and detection failure ratio of 3％ and excellent tracking.

A neural network model for the perception of occluded surfaces from subjective contours

A neural network model for the perception of occluded surfaces from subjective contours has been presented. This model employs an important two-stage process of the induced stimuli extraction system (ISES) and subjective surfaces perception system (SSPS). The former system extracts the induced stimuli for the perception of subjective surfaces, and the latter forms the subjective surfaces from the induced stimuli. The proposed model is based on the mechanism of feature extraction found in the visual pathway. The results of the experiment showed that the proposed model was successful not only in extracting the induced stimuli for the perception of subjective contours, but also in perceiving the subjective surface from the induced stimuli.

Concentration analysis by detecting face features of learners

The paper presents an analysis on the concentration of learning. By capturing video images of students, the proposed method detects and analyzes facial features from the image data and determines the state of learners concentration. Since the concentration is important to the learners, this method is applied to the classrooms. First, feature points are generated from the face and then feature points of the face are used to determine non-focused state. The length of the front face is used to make a decision for the face change. The coordinate value of the facial center is used to decide the face turns. The criteria value of the opened eye is used to decide whether the closed eyes or the opened eyes. Through the experiments, the proposed method detects the concentration up to 90%.

A Study of Post-processing Methods of Clustering Algorithm and Classification of the Segmented Regions

Some clustering algorithms have a problem that an image is over-segmented since both the spatial information between the segmented regions is not considered and the number of the clusters is defined in advance. Therefore, they are difficult to be applied to the applicable fields. This paper proposes the new post-processing methods, a reclassification of the inhomogeneous clusters and a region merging using Baysian algorithm, that improve the segmentation results of the clustering algorithms. The inhomogeneous cluster is firstly selected based on variance and between-class distance and it is then reclassified into the other clusters in the reclassification step. This reclassification is repeated until the optimal number determined by the minimum average within-class distance. And the similar regions are merged using Baysian algorithm based on Kullbeck-Leibler distance between the adjacent regions. So we can effectively solve the over-segmentation problem and the result can be applied to the applicable fields. Finally, we design a classification system for the segmented regions to validate the proposed method. The segmented regions are classified by SVM(Support Vector Machine) using the principal colors and the texture information of the segmented regions. In experiment, the proposed method showed the validity for various real-images and was effectively applied to the designed classification system.