scispace - formally typeset
Search or ask a question
Author

Syed Mohd Zahid Syed Zainal Ariffin

Bio: Syed Mohd Zahid Syed Zainal Ariffin is an academic researcher from Universiti Teknologi MARA. The author has contributed to research in topics: Image fusion & Euclidean distance. The author has an hindex of 3, co-authored 8 publications receiving 17 citations. Previous affiliations of Syed Mohd Zahid Syed Zainal Ariffin include Asia Pacific University of Technology & Innovation.

Papers
More filters
Proceedings ArticleDOI
01 May 2017
TL;DR: Five fusion methods, namely, simple average, mean average, average discrete wavelet transform (DWT), optimized DWT and weighted DWT were tested to study whether thermal and visible image fusion improves ear recognition given illumination variations.
Abstract: This paper intends to study whether thermal and visible image fusion improves ear recognition given illumination variations. Five fusion methods, namely, simple average, mean average, average discrete wavelet transform (DWT), optimized DWT and weighted DWT were tested. The experiments were done using DIAST Variability Illuminated Thermal and Visible Ear Image Datasets. There were two experiments conducted. The first experiment is done using images with three different illumination conditions (i.e. dark, average and bright illuminations) while the second experiment only considers average and brightly illuminated image. The evaluation was done based on ear recognition accuracy where features were extracted based on the histogram of oriented gradients (HOG) while support vector machine (SVM) was used as the classifier. In Experiment 1, thermal images performed best with 96.36% recognition rate while visible images was the lowest with 62.73% accuracy. In Experiment 2, all three DWT fusions scored 95.45% accuracy, surpassing thermal image (94.55%) while visible image still is the lowest with 90.00% recognition rate.

11 citations

Journal ArticleDOI
TL;DR: This paper proposes a novel CNN model that can perform equally well for very underexposed or overexposed images or known as uniform illumination invariant and shows that 100% accuracy is achieved for images with lumens ranging above 30 but having problem with Lumens less than 10 lux.
Abstract: Current deep convolution neural network (CNN) has shown to achieve superior performance on a number of computer vision tasks such as image recognition, classification and object detection. The deep network was also tested for view-invariance, robustness and illumination invariance. However, the CNN architecture has thus far only been tested on non-uniform illumination invariant. Can CNN perform equally well for very underexposed or overexposed images or known as uniform illumination invariant? This is the gap that we are addressing in this paper. In our work, we collected ear images under different uniform illumination conditions with lumens or lux values ranging from 2 lux to 10,700 lux. A total of 1,100 left and right ear images from 55 subjects are captured under natural illumination conditions. As CNN requires considerably large amount of data, the ear images are further rotated at every 5o angles to generate 25,300 images. For each subject, 50 images are used as validation/testing dataset, while the remaining images are used as training datasets. Our proposed CNN model is then trained from scratch and validation and testing results showed recognition accuracy of 97%. The results showed that 100% accuracy is achieved for images with lumens ranging above 30 but having problem with lumens less than 10 lux

11 citations

Proceedings ArticleDOI
01 Sep 2016
TL;DR: This paper introduces ear image datasets that were captured in both visible and thermal domain, and presents the detailed process of image acquisition, post-processing, and preparation of the ear datasets.
Abstract: As the research in ear biometrics grows, there is a need to have a publicly available ear images database for researchers to test and validate their methods. There are many available ear databases captured in visible light domain. However, images in those databases are mostly acquired in a controlled environment, not in naturally illuminated environments. Up to the point of writing, there is no thermalbased ear image dataset available for research usage. In this paper, we introduce our ear image datasets that were captured in both visible and thermal domain. We present the detailed process of image acquisition, post-processing, and preparation of the ear datasets. The ear dataset consists of 2,200 images captured from 55 subjects in five different illumination conditions, ranging from 2 lux to 10,700 lux, measured using lux meter.

9 citations

Journal Article
TL;DR: The possibility of fusing thermal and visible images to improve ear recognition ability in images of varying illuminations is discussed and image fusion is proposed to accentuate the strengths of both spectra for ear recognition.
Abstract: This paper discusses the possibility of fusing thermal and visible images to improve ear recognition ability in images of varying illuminations. Since thermal image is known to remain invariant to lighting changes and visible images are able to capture feature details, image fusion is proposed to accentuate the strengths of both spectra for ear recognition. Two popular image fusion techniques, weighted average (WA) and discrete wavelet transform (DWT) are used as a preliminary investigation. Eigenvectors are extracted from the fused image and recognition is performed using metric distance measure. With 67.5% recognition rate, DWT fused images performed better than WA fused images (63.75%). Thermal images, on the other hand, achieved 68.75% recognition rate. Even though thermal images performed slightly better than DWT fused images by 1.25%, the difference is deemed as insignificance due to the small dataset used and the primitive fusion rules employed. Further studies on the fusion techniques need to be done to improve fusion method.

3 citations

Proceedings ArticleDOI
01 Aug 2014
TL;DR: A cross-band ear recognition to overcome the variant illumination problem and determine the individual identity (intra- and inter-variance) of the ear region using Euclidean distance.
Abstract: Ear biometric is slowly gaining its position in biometric studies. Just like fingerprint and iris, the ears are unique and have other advantages over current regular biometric methods. Besides those advantages, there are some issues arising for ear recognition. One of those is regarding the illumination. Low illumination may result in low quality image acquired resulting in low recognition rate. Based on this situation, we proposed a cross-band ear recognition to overcome the variant illumination problem. This method starts by measuring the environments illumination which will determine which type of images (i.e.: thermal or visible) acquired to be processed. Once determined, the images will undergo pre-processing before the ear region is being localized using Viola-Jones approach with Haar-like feature. The ear features will be extracted using local binary patterns operator. Euclidean distance of the feature of test image and database images will be calculated. The lowest Euclidean value will determine the individual identity (intra- and inter-variance).

1 citations


Cited by
More filters
Proceedings ArticleDOI
14 Jun 2020
TL;DR: This paper presents a visible and infrared image fusion benchmark (VIFB), identifies effective algorithms for robust image fusion and gives some observations on the status and future prospects of this field.
Abstract: Visible and infrared image fusion is an important area in image processing due to its numerous applications. While much progress has been made in recent years with efforts on developing image fusion algorithms, there is a lack of code library and benchmark which can gauge the state-of-the-art. In this paper, after briefly reviewing recent advances of visible and infrared image fusion, we present a visible and infrared image fusion benchmark (VIFB) which consists of 21 image pairs, a code library of 20 fusion algorithms and 13 evaluation metrics. We also carry out extensive experiments within the benchmark to understand the performance of these algorithms. By analyzing qualitative and quantitative results, we identify effective algorithms for robust image fusion and give some observations on the status and future prospects of this field.

95 citations

Proceedings ArticleDOI
26 Mar 2018
TL;DR: This paper applies Transfer Learning to the well-known AlexNet Convolution Neural Network (AlexNet CNN) for human recognition based on ear images and fine-tuned AlexNet CNN to suit the problem domain.
Abstract: Transfer Learning is an efficient approach of solving classification problem with little amount of data In this paper, we applied Transfer Learning to the well-known AlexNet Convolution Neural Network (AlexNet CNN) for human recognition based on ear images We adopted and fine-tuned AlexNet CNN to suit our problem domain The last fully connected layer is replaced with another fully connected layer to recognize 10 classes instead of 1000 classes Another Rectified Linear Unit (ReLU) layer is also added to improve the non-linear problem-solving ability of the network To train the fine-tuned network, we allocate 250 ear images taken from 10 subjects for training, and 50 ear images are used for validation and testing The proposed fine-tuned network works well in our application as we get 100% validation accuracy

83 citations

Journal ArticleDOI
TL;DR: Zhang et al. as mentioned in this paper proposed a region-free object detector named YOLO-FIR for infrared (IR) images by compressing channels, optimizing parameters, etc.
Abstract: To solve object detection issues in infrared images, such as a low recognition rate and a high false alarm rate caused by long distances, weak energy, and low resolution, we propose a region-free object detector named YOLO-FIR for infrared (IR) images with YOLOv5 core by compressing channels, optimizing parameters, etc. An improved infrared image object detection network, YOLO-FIRI, is further developed. Specifically, while designing the feature extraction network, the cross-stage-partial-connections (CSP) module in the shallow layer is expanded and iterated to maximize the use of shallow features. In addition, an improved attention module is introduced in residual blocks to focus on objects and suppress background. Moreover, multiscale detection is added to improve small object detection accuracy. Experimental results on the KAIST and FLIR datasets show that YOLO-FIRI demonstrates a qualitative improvement compared with the state-of-the-art detectors. Compared with YOLOv4, the mean average precision (mAP50) of YOLO-FIRI is increased by 21% on the KAIST dataset, the speed is reduced by 62%, the parameters are decreased by 89%, the weight size is reduced by more than 94%, and the computational costs are reduced by 84%. Compared with YOLO-FIR, YOLO-FIRI has an approximately 5% to 20% improvement in AP, AR (average recall), mAP50, F1, and mAP50:75. Furthermore, due to the shortcomings of high noise and weak features, image fusion can be applied to image preprocessing as a data enhancement method by fusing visible and infrared images based on a convolutional neural network.

67 citations

Journal ArticleDOI
TL;DR: The entire process of ear recognition, including detection, preprocessing, unimodal recognition including feature extraction and decision of classification or matching, and multimodal Recognition based on inter-level and intra-level fusion are classified comprehensively.
Abstract: As one of the most important biometrics, ear biometrics is getting more and more attention. Ear recognition has unique advantages and can make identification more secure and reliable together with other biometrics (e.g. face and fingerprint). Therefore, we investigate related information about ear recognition and classify the entire process of ear recognition, including detection, preprocessing, unimodal recognition including feature extraction and decision of classification or matching, and multimodal recognition based on inter-level and intra-level fusion. Unimodal and multimodal recognition are proposed comprehensively. In addition, inter-level and intra-level fusion are divided into different fusion ways. At the same time, we compare recognition results under the same dataset and analyze the difficulty of some datasets. In the end, challenges and outlook of ear recognition are also mentioned to expect to provide readers with some help about future directions and problems that should be overcome.

18 citations