Md. Al-Amin Bhuiyan

Face recognition using Gabor Filters

Gabor based face representation has achieved enormous success in face recognition. This paper addresses a novel algorithm for face recognition using neural networks trained by Gabor features. The system is commences on convolving some morphed images of particular face with a series of Gabor filter co-efficient at different scales and orientations. Two novel contributions of this paper are: scaling of RMS contrast, and contribution of morphing as an advancement of image recognition perfection. The neural network employed for face recognition is based on the multy layer perceptron (MLP) architecture with back-propegation algorithm and incorporates the convolution filter response of Gabor jet. The effectiveness of the algorithm has been justified over a morphed facial image database with images captured in different illumination conditions.

Transfer between pose and illumination training in face recognition.

The relationship between pose and illumination learning in face recognition was examined in a yes-no recognition paradigm. The authors assessed whether pose training can transfer to a new illumination or vice versa. Results show that an extensive level of pose training through a face-name association task was able to generalize to a new illumination (Experiments 1 and 3), but an equal level of illumination training failed to generalize to a new pose (Experiment 2). The transfer of pose training was likely to depend on a relatively extensive level of training because the same faces with reduced level of exposure (Experiment 4) were unable to reproduce the transfer effect. The findings suggest that generalization of pose training may be extended to different types of image variation, whereas generalization of illumination training may be confined within the trained type of variation.

Person to Camera Distance Measurement Based on Eye-Distance

This paper presents a novel person to camera distance measuring system based on eye-distance. The distance between centers of two eyes is used for measuring the person to camera distance. The variation in eye-distance (in pixels) with the changes in camera to person distance (in inches) is used to formulate the distance measuring system. The system starts with computing the distance between two eyes of a person and then person to camera distance is measured. The proposed distance measurement system is relatively simple and inexpensive to implement as it does not require any other external distance measuring tools. Experimental results show the effectiveness of the system with an average accuracy of 94.11%.

Face and gesture recognition using subspace method for human-robot interaction

This paper presents a vision-based face and gesture recognition system for human-robot interaction. By using subspace method, face and predefined hand poses are detected from the three largest skin-like regions that are segmented using YIQ color representation system. In this subspace method we consider separate eigenspaces for each class or pose. Gesture is recognized using the rule-based approach whenever the combination of three skin-like regions at a particular image frame matches with the predefined gesture. These gesture commands are sent to robot through TCP/IP network for human-robot interaction. Using subspace method pose invariant face recognition has also been addressed. The effectiveness of this method has been demonstrated by interacting with an entertainment robot named AIBO.

Identification of actors drawn in Ukiyoe pictures

Abstract This paper presents the development of line image keywords for the identification of actors drawn in Japanese traditional painting pictures known as Ukiyoe pictures. The system is based on visual features of the face from the image database files and is organized as a set of classifiers whose outputs are integrated after a normalization step. Line profile from the picture has been extracted in this investigation and has been approximated by Bezier curves. A learning algorithm has been developed to obtain the control points at high accuracy. A new curve matching method has been developed based on the feature points, rather than the corresponding points. This method can automatically fit a set of data points with piecewise geometrically continuous third order Bezier curves. Last of all, a new approach for distance calculation, namely “apple-node distance” has been introduced here for similarity calculation in image retrieval systems. The computation of similarity between curves has been established on the basis of this “apple-node” distance. The effectiveness of our method has been confirmed through computer simulation. The method developed here can be expanded to one of three dimensional shape-analyzing tools.

Digital image enhancement with fuzzy rule-based filtering

Image enhancement is a technique to improve the quality of an image. The aim of image enhancement technique is to improve the interpretability or perception of information in images for human viewers, or to provide better input for other automated image processing techniques. This paper presents a new approach for image enhancement with fuzzy rule-based filtering. Compared to other non-linear techniques, fuzzy filter gives the better performance and is able to represent knowledge in a comprehensible way.

ART Network based Face Recognition with Gabor Filters

Gabor-based face representation has achieved enormous success in face recognition. This research addresses a hybrid neural network solution for face recognition trained with Gabor features. The system is commenced on convolving a face image with a series of Gabor filter coefficients at different scales and orientations. The neural network employed for face recognition is based on BAM for dimensional reduction and multi-layer perception with backpropagation algorithm for training the Gabor features. The effectiveness of the algorithm has been justified over a face database with images captured at different illumination conditions.

A novel approach of image morphing based on pixel transformation

Image morphing has been the subject of much attention in recent years. It has proven to be a powerful tool for visual effects in film and television, depicting the fluid transformation of one digital image into another. This paper presents a new approach of image morphing based on pixel transformation that depicts the transformation of pixels with their neighborhoods. The method is organized with the replacement of the pixel values of a source image and convolving the neighbor with the help of a mask. This algorithm changes an image from a particular side or from the center. Experimental results demonstrate that the method is fast and efficient for image morphing.

Neural Network based Road Sign Recognition

A recent surge of interest is to recognize Road Signs. Signs are visual languages that represent some special circumstantial information of environment. They provide important information for guiding, warning people to make their movements safer, easier and more convenient. This thesis presents a hybrid neural network solution for Road sign recognition which combines local image sampling and artificial neural network. The method is based on BAM for dimensional reduction and multi-layer perception with backpropagation algorithm has been used for training the network. It has been found from practical observations that the number of iterations required to train the network is enormous. The capability of recognition of a neural network increases with increasing the training accuracy. For this process each sign is converted to a designated M×N feature matrix. These feature matrices of signs are then fed into the neural network as input patterns. The neural network is trained with the set of input patterns of the digits to acquire separate knowledge corresponding to each Road sign. In order to justify the effectiveness of the system, different test patterns of the signs are used to verify the system. Experimental results demonstrate that the system is capable of recognizing Road signs with 98% accuracy.

Gesture Recognition for Human-Robot Interaction Through a Knowledge Based Software Platform

The task of real-time hand gesture recognition is extremely challenging due to a number of DOFs of hand pose and motion. However, for human-robot interaction in natural ways, gesture can provide a powerful interface tool for commanding a robot to perform a specific task. This paper presents a vision-based real-time gesture recognition system by segmenting the three largest skin color components and template-matching techniques with multiple features. Gesture commands are generated whenever the combinations of three skin-like regions at a particular image match with the predefined gestures. These gesture commands are sent to robots through a knowledge based software platform for human-robot interaction. The effectiveness of our method has been demonstrated by interacting with an entertainment robot named AIBO.