Asuman Günay

Automatic age classification with LBP

Estimating the age exactly and then producing the younger and older images of the person is important in security systems design. In this paper local binary patterns are used to classify the age from facial images. The local binary patterns (LBP) are fundamental properties of local image texture and the occurrence histogram of these patterns is an effective texture feature for face description. In the study we classify the FERET images according to their ages with 10 years intervals. The faces are divided into small regions from which the LBP histograms are extracted and concatenated into a feature vector to be used as an efficient face descriptor. For every new face presented to the system, spatial LBP histograms are produced and used to classify the image into one of the age classes. In the classification phase, minimum distance, nearest neighbor and k-nearest neighbor classifiers are used. The experimental results have shown that system performance is 80% for age estimation.

Automatic Detection of Anthropometric Features from Facial Images

In this study we try to detect anthropometric features and estimate age from facial images. We designed a neural network and trained it for face detection. In order to locate the facial features we calculate the vertical and horizontal projections and search them for minimums and maximums. Later we calculate some geometrical ratios and differences which are used for age estimation. Experimental results show that our algorithm can detect the face and facial features successfully.

Age Estimation Based on Local Radon Features of Facial Images

This paper proposes a new age estimation method relying on regional Radon features of facial images and regression. Radon transform converts a pixel represented image an equivalent, lower dimensional and more geometrically informative Radon pixel image and it brings a large advantage achieving global geometric affine invariance. Proposed method consists of four modules: preprocessing, feature extraction with Radon transform, dimensionality reduction with PCA and age estimation with multiple linear regression. We conduct our experiments on FG-NET, MORPH and FERET databases and the results have shown that proposed method has better results than many conventional methods on all databases.

Age Estimation Based on Hybrid Features of Facial Images

This paper proposes a new age estimation method using hybrid features produced by fusing the global and local features of facial images at decision level. The global facial features are extracted using active appearance models (AAM) which contains both the shape and appearance of a human face. In the local feature extraction phase, the wrinkle features are extracted using a set of Gabor filters, capable of extracting deep and fine wrinkles in different directions and the skin features are extracted using local binary patterns (LBP), capable of extracting the detailed textures of skin. After the feature extraction module, three aging functions are modeled separately with multiple linear regression. Then decision level fusion is performed to combine the results of these aging functions to make a final decision. Experimental results showed that the performance of the proposed method was superior to that of the previous methods when using the FG-NET and PAL aging databases.

Age Estimation Based on AAM and 2D-DCT Features of Facial Images

This paper proposes a novel age estimation method - Global and Local feAture based Age estiMation (GLAAM) - relying on global and local features of facial images. Global features are obtained with Active Appearance Models (AAM). Local features are extracted with regional 2D-DCT (2- dimensional Discrete Cosine Transform) of normalized facial images. GLAAM consists of the following modules: face normalization, global feature extraction with AAM, local feature extraction with 2D-DCT, dimensionality reduction by means of Principal Component Analysis (PCA) and age estimation with multiple linear regression. Experiments have shown that GLAAM outperforms many methods previously applied to the FG-NET database.

A new facial age estimation method using centrally overlapped block based local texture features

This paper introduces a new age estimation method based on the fusion of local features extracted using histogram-based local texture descriptors. In the study the age estimation performances of well-known powerful texture descriptor Local Binary Patterns (LBP), and new texture descriptors Weber Local Descriptor (WLD) and Local Phase Quantization (LPQ) which have not been analyzed in depth for age estimation, are investigated. Also multi-scale and spatial texture analysis is performed for all descriptors. In the spatial texture analysis, a new approach using the Centrally Overlapped Blocks (COB) obtained by combining the centers of discrete blocks is proposed to capture the related information between the blocks. Then feature fusion is performed to investigate the age estimation accuracies of different combinations of local texture descriptors. After dimensionality reduction with Principal Component Analysis (PCA), Multiple Linear Regression (MLR) is used to estimate the specific age. The results show that the age estimation accuracy of the proposed method is better when compared to previous methods on FG-NET, MORPH and PAL databases.

Facial age estimation using spatial Weber Local Descriptor

This paper introduces a novel age estimation method using a new texture descriptor Weber Local Descriptor (WLD). This texture descriptor is analyzed in depth for age estimation problem. In the study, the multi-scale versions of holistic and spatial WLD (SWLD) descriptors are used to extract the age related features from normalized facial images. After finding a lower dimensional feature subspace, age estimation is performed using multiple linear regression. In addition a new approach of dividing image into regions for spatial texture extraction is proposed. Experiments on FG-NET, MORPH and PAL databases have shown that the proposed method gives better accuracy than the state of art age estimation approaches.

Shredded banknotes reconstruction using AKAZE points

Shredded banknote reconstruction is a recent topic and can be viewed as solving large-scale jigsaw puzzles. Also, problems such as reconstruction of fragmented documents, photographs and historical artefacts are closely related with this topic. The high computational complexity of these problems increases the need for the development of new methods Reconstruction of shredded banknotes consists of three main stages. (1) Matching fragments with a reference banknote. (2) Aligning the fragments by rotating at certain angles. (3) Assembling the fragments. The existing methods can successfully applied to synthetic banknote fragments which are created in computer environment. But when real banknote reconstruction problem is considered, different sub problems arise and make the existing methods inadequate. In this study, a keypoint based method, named AKAZE, was used to make the matching process effective. This is the first study that uses the AKAZE method in the reconstruction of shredded banknotes. A new method for fragment alignment has also been proposed. In this method, the convex hulls that contain all true matched AKAZE keypoints were found on reference banknote and fragments. The orientations of fragments were estimated accurately by comparing these convex polygons. Also, a new criterion was developed to reveal the success rates of reconstructed banknotes. In addition, two different data sets including real and synthetic banknote fragments of different countries were created to test the success of proposed method.

Investigating the Effects of Facial Regions to Age Estimation

Aging process causes evident alterations on human facial appearance. Real world age progression on human face is personalized and related with many factors such as, genetics, living style, eating habits, facial expressions, climate etc. The wide degree of variations on facial appearance of different individuals affects the age estimation performance. In accordance with these facts discovering the aging information contained in facial regions is an important issue in automatic age estimation. Thus the facial regions emphasizing the aging information can be used for more accurate age estimation. In this context, age estimation performances of facial regions (eye, nose, mouth and chin, cheeks and sides of mouth) are investigated in this paper. For this purpose, an age estimation method is designed to produce an estimate of the age of a subject by using the texture features extracted from facial regions. In this method the facial images are warped into the mean shape thus variations of head pose and scale are eliminated and the texture information of facial images are aligned. Then the holistic and spatial texture features are extracted from facial regions using Local Phase Quantization (LPQ) texture descriptor, robust to blur, illumination and expression variations. After the low dimensional representation of these features, a linear aging function is learned using multiple linear regression. In the experiments FGNET and PAL databases are used to evaluate the age estimation accuracies of facial regions i.e. eye, nose, mouth and chin, cheek and sides of mouth, separately. The results have shown that the eye region carries the most significant information for age estimation. Also the mouth and chin, cheek regions are effective in the prediction of age. The results also have shown that, using the spatial texture features enhances the discriminative power of the texture descriptor and thus increases the estimation accuracy.

Age estimation using hybrid features of facial images

This paper presents a novel age estimation method using hybrid features, which are the combination of global and local features of facial images. Global and local features are extracted using Active Appearance Models and Discrete Cosine Transform, respectively. Then these features are fused in the feature level, and the age is estimated using multiple linear regression. Experiments on FG-NET database have shown that, the hybrid features improves the age estimation performance.