Pawanesh Abrol
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Pawanesh Abrol has a total of 25 co-authors. Here are the top co-authors by number of shared publications:
- Palak Mahajan
Central University of Jammu
4 shared publications - Mohsina Ishrat
Department of Computer Science & IT, University of Jammu (J&K), Jammu, India
2 shared publications - Parveen Kumar Lehana
University of Jammu
2 shared publications - Anjana Sharma
GGM Science College
1 shared publication - Anjana Sharma1 shared publication
- Arvind Selwal
Department of Computer Science & Information Technology, Central University of Jammu, Jammu, India
1 shared publication
Latest Publications
- 15 January 2026
- Published by Springer Nature
- p. 461-473
Performance Assessment of Automatic Image Caption Generation via Bilingual Evaluation Understudy Protocol
- 15 January 2026
- Published by Springer Nature
- p. 461-473
Performance Assessment of Automatic Image Caption Generation via Bilingual Evaluation Understudy Protocol
- 17 December 2025
- Published by World Scientific Pub Co Pte Ltd in International Journal of Image and Graphics
A Systematic Review of Image Quality Assessment Methods: Hitherto and Futuristic Perspectives
Digital images are the most frequently generated and communicated media from a diverse range of applications, such as medical imaging, satellite imagery, social media, secured surveillance, robotics, computer vision, etc. The image quality is one of the vital characteristics that is essential for computer vision tasks, where key features play a significant role in object detection and classification. Usually, images are subjected to several inconsistencies, such as noise, blurring, resolution, intensity variations, distortions, etc. Thus, image quality assessment (IQA) and enhancement are imperative for accurate and efficient machine learning models in computer vision applications. This study aims to provide an in-depth review and analysis of existing state-of-the-art (SOTA) IQA techniques. Besides, a new taxonomy is presented to classify various IQA techniques, as well as various evaluation protocols (i.e. metrics and benchmark datasets) that are frequently used to assess these methods. However, noteworthy progress has been observed in IQA techniques, but there exist several research challenges in this active field of digital imaging. One of the crucial challenges is to deploy modern deep learning models for IQA and enhancement, as these require comparatively larger datasets. The quality assessment of generative AI-based images is an additional challenge, as they are subjected to several quality issues. Moreover, multi-spectral and super-resolution image quality assessment is another open challenge that requires the attention of investigators in this field of research.
- 17 December 2025
- Published by World Scientific Pub Co Pte Ltd in International Journal of Image and Graphics
A Systematic Review of Image Quality Assessment Methods: Hitherto and Futuristic Perspectives
Digital images are the most frequently generated and communicated media from a diverse range of applications, such as medical imaging, satellite imagery, social media, secured surveillance, robotics, computer vision, etc. The image quality is one of the vital characteristics that is essential for computer vision tasks, where key features play a significant role in object detection and classification. Usually, images are subjected to several inconsistencies, such as noise, blurring, resolution, intensity variations, distortions, etc. Thus, image quality assessment (IQA) and enhancement are imperative for accurate and efficient machine learning models in computer vision applications. This study aims to provide an in-depth review and analysis of existing state-of-the-art (SOTA) IQA techniques. Besides, a new taxonomy is presented to classify various IQA techniques, as well as various evaluation protocols (i.e. metrics and benchmark datasets) that are frequently used to assess these methods. However, noteworthy progress has been observed in IQA techniques, but there exist several research challenges in this active field of digital imaging. One of the crucial challenges is to deploy modern deep learning models for IQA and enhancement, as these require comparatively larger datasets. The quality assessment of generative AI-based images is an additional challenge, as they are subjected to several quality issues. Moreover, multi-spectral and super-resolution image quality assessment is another open challenge that requires the attention of investigators in this field of research.
- Open Access
- 1 December 2024
- Published by Institute of Advanced Engineering and Science in International Journal of Informatics and Communication Technology (IJ-ICT)
- Vol. 13 (3) , 333-343
Thermal imaging-based identification of facial features in noisy environment
Face identification is amongst the most efficacious and extensive applications in biometrics involving extraction and locating facial features. With identification being monotonous task attributable to reliance on parameters like varied cameras, fluctuating backgrounds, and exposure to the environment in which an individual is present. Thermal imaging is endeavoring to resolve the accuracy issue of apparent imaging, such as lighting and brightness intensity, among all biometric variables. This paper presents a study of thermal imaging and effective methods involved in the feature extraction process for facial features with thermal imaging under the influence of varied noise. A novel face dataset is created TID comprising 27 thermal images and its corresponding visual band image using Fluke 480 Ti Pro camera. The study analyses detection efficiency of six feature extraction techniques in visible and thermal bands in facial features identification. Also, the influence of noise in the thermal band within the region of interest using feature points FIN, FOUT has been estimated. Throughout TID dataset, ORB extraction technique has been able to identify strongest inlier features FIN to a maximum extent with detection around the nose, eyes, and mouth. Further, results indicate feature detection in thermal images being invariant to effect of noise for detecting facial features.
- 1 December 2024
- Published by Institute of Advanced Engineering and Science in International Journal of Informatics and Communication Technology (IJ-ICT)
- Vol. 13 (3) , 333-343
Open AccessThermal imaging-based identification of facial features in noisy environment
Face identification is amongst the most efficacious and extensive applications in biometrics involving extraction and locating facial features. With identification being monotonous task attributable to reliance on parameters like varied cameras, fluctuating backgrounds, and exposure to the environment in which an individual is present. Thermal imaging is endeavoring to resolve the accuracy issue of apparent imaging, such as lighting and brightness intensity, among all biometric variables. This paper presents a study of thermal imaging and effective methods involved in the feature extraction process for facial features with thermal imaging under the influence of varied noise. A novel face dataset is created TID comprising 27 thermal images and its corresponding visual band image using Fluke 480 Ti Pro camera. The study analyses detection efficiency of six feature extraction techniques in visible and thermal bands in facial features identification. Also, the influence of noise in the thermal band within the region of interest using feature points FIN, FOUT has been estimated. Throughout TID dataset, ORB extraction technique has been able to identify strongest inlier features FIN to a maximum extent with detection around the nose, eyes, and mouth. Further, results indicate feature detection in thermal images being invariant to effect of noise for detecting facial features.
- Open Access
- 1 August 2023
- Published by Institute of Advanced Engineering and Science in International Journal of Informatics and Communication Technology (IJ-ICT)
- Vol. 12 (2) , 115-126
Analyzing performance of deep learning models under the presence of distortions in identifying plant leaf disease
Convolutional neural networks (CNN) trained using deep learning (DL) have advanced dramatically in recent years. Researchers from a variety of fields have been motivated by the success of CNNs in computer vision to develop better CNN models for use in other visually-rich settings. Successes in image classification and research have been achieved in a wide variety of domains throughout the past year. Among the many popularized image classification techniques, the detection of plant leaf diseases has received extensive research. As a result of the nature of the procedure, image quality is often degraded and distortions are introduced during the capturing of the image. In this study, we look into how various CNN models are affected by distortions. Corn-maze leaf photos from the 4,188-image corn or maize leaf Dataset (split into four categories) are under consideration. To evaluate how well they handle noise and blur, researchers have deployed pre-trained deep CNN models like visual geometry group (VGG), InceptionV3, ResNet50, and EfficientNetB0. Classification accuracy and metrics like as recall and f1-score are used to evaluate CNN performance.
- 1 August 2023
- Published by Institute of Advanced Engineering and Science in International Journal of Informatics and Communication Technology (IJ-ICT)
- Vol. 12 (2) , 115-126
Open AccessAnalyzing performance of deep learning models under the presence of distortions in identifying plant leaf disease
Convolutional neural networks (CNN) trained using deep learning (DL) have advanced dramatically in recent years. Researchers from a variety of fields have been motivated by the success of CNNs in computer vision to develop better CNN models for use in other visually-rich settings. Successes in image classification and research have been achieved in a wide variety of domains throughout the past year. Among the many popularized image classification techniques, the detection of plant leaf diseases has received extensive research. As a result of the nature of the procedure, image quality is often degraded and distortions are introduced during the capturing of the image. In this study, we look into how various CNN models are affected by distortions. Corn-maze leaf photos from the 4,188-image corn or maize leaf Dataset (split into four categories) are under consideration. To evaluate how well they handle noise and blur, researchers have deployed pre-trained deep CNN models like visual geometry group (VGG), InceptionV3, ResNet50, and EfficientNetB0. Classification accuracy and metrics like as recall and f1-score are used to evaluate CNN performance.
- 6 March 2023
- Published by Springer Nature in Neural Computing and Applications
- Vol. 35 (17) , 12675-12689
Wrapper-based optimized feature selection using nature-inspired algorithms
- 6 March 2023
- Published by Springer Nature in Neural Computing and Applications
- Vol. 35 (17) , 12675-12689
Wrapper-based optimized feature selection using nature-inspired algorithms
- 1 January 2023
- Published by ACS Publisher in Applied Biological Research
- Vol. 25 (3) , 361-371
Similar Plant Species Classification with Dual Input Transfer Learning Models
The classification of plant species is crucial for preserving biodiversity, identifying endangered species, preventing diseases, and controlling weeds. However, with the vast number of species and their similar physical appearances, identifying and classifying them correctly can be difficult for human experts. Artificial Intelligence, such as Convolution Neural Networks (CNNs), can aid in automatic plant species recognition, especially for identifying similar species that require the learning of subtle differences. In the present investigation, transfer learning models such as MobileNet, AlexNet, and GoogLeNet were used to classify ten citrus species that look alike. A dataset of images of fruits and leaves of these citrus plants was compiled, and all images were segmented and their backgrounds were subtracted to create a new segmented dataset. In addition, instead of using only one organ as input to the CNN, dual inputs of both fruits and leaves were also used. The classification accuracy of 78.25% was achieved when MobileNet model was used on original dataset of fruits and it extended to 97.25% when GoogLeNet model was used on segmented dataset when both the organs were used as input. Evidently, present study provides innovative methods and techniques to accurately distinguish and classify visually similar plant species.
- 1 January 2023
- Published by ACS Publisher in Applied Biological Research
- Vol. 25 (3) , 361-371
Similar Plant Species Classification with Dual Input Transfer Learning Models
The classification of plant species is crucial for preserving biodiversity, identifying endangered species, preventing diseases, and controlling weeds. However, with the vast number of species and their similar physical appearances, identifying and classifying them correctly can be difficult for human experts. Artificial Intelligence, such as Convolution Neural Networks (CNNs), can aid in automatic plant species recognition, especially for identifying similar species that require the learning of subtle differences. In the present investigation, transfer learning models such as MobileNet, AlexNet, and GoogLeNet were used to classify ten citrus species that look alike. A dataset of images of fruits and leaves of these citrus plants was compiled, and all images were segmented and their backgrounds were subtracted to create a new segmented dataset. In addition, instead of using only one organ as input to the CNN, dual inputs of both fruits and leaves were also used. The classification accuracy of 78.25% was achieved when MobileNet model was used on original dataset of fruits and it extended to 97.25% when GoogLeNet model was used on segmented dataset when both the organs were used as input. Evidently, present study provides innovative methods and techniques to accurately distinguish and classify visually similar plant species.
- 14 February 2022
- Published by Springer Nature in Multimedia Tools and Applications
- Vol. 81 (7) , 10007-10022
Visual search analysis using parametric fixations
- 14 February 2022
- Published by Springer Nature in Multimedia Tools and Applications
- Vol. 81 (7) , 10007-10022
Visual search analysis using parametric fixations
- 11 March 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 17 (12) , 8046-8056
Perceptual Quality Evaluation of Hazy Natural Images
Haze is an intrusion element that disrupts color fidelity and contrast of outdoor natural images, affecting their perceptual quality. The differential characteristics of hazy images compared to other natural images restrict the generalization of existing image quality assessment (IQA) algorithms. At the same time, efficient IQA algorithms for predicting the perceptual quality of naturally hazed images have not been proposed in the literature due to lack of a relevant dataset. To address this, we build the IIT-JMU Hazy Image Dataset comprising of 1000 high-definition hazy natural images consisting of diverse categories such as landscape, forests, roads, seascapes, and cityscapes, along with their subjective quality ratings. We present an analysis of existing natural-scene-statistics-based IQA algorithms on hazy natural images. In this article, we propose a convolutional-neural-network-based quality assessment algorithm for hazy natural images along with an IQA metric called deep learning-based haze perceptual quality evaluator (DLHPQE). The proposed DLHPQE efficiently predicts the perceptual quality of hazy natural images without a reference. Our results demonstrate that the DLHPQE outperforms existing state-of-the-art no-reference IQAs in terms of several performance parameters such as Pearson linear correlation coefficient, Spearman rank-order correlation coefficient, Kendall's rank-order correlation coefficient, and root-mean-square error.
- 11 March 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 17 (12) , 8046-8056
Perceptual Quality Evaluation of Hazy Natural Images
Haze is an intrusion element that disrupts color fidelity and contrast of outdoor natural images, affecting their perceptual quality. The differential characteristics of hazy images compared to other natural images restrict the generalization of existing image quality assessment (IQA) algorithms. At the same time, efficient IQA algorithms for predicting the perceptual quality of naturally hazed images have not been proposed in the literature due to lack of a relevant dataset. To address this, we build the IIT-JMU Hazy Image Dataset comprising of 1000 high-definition hazy natural images consisting of diverse categories such as landscape, forests, roads, seascapes, and cityscapes, along with their subjective quality ratings. We present an analysis of existing natural-scene-statistics-based IQA algorithms on hazy natural images. In this article, we propose a convolutional-neural-network-based quality assessment algorithm for hazy natural images along with an IQA metric called deep learning-based haze perceptual quality evaluator (DLHPQE). The proposed DLHPQE efficiently predicts the perceptual quality of hazy natural images without a reference. Our results demonstrate that the DLHPQE outperforms existing state-of-the-art no-reference IQAs in terms of several performance parameters such as Pearson linear correlation coefficient, Spearman rank-order correlation coefficient, Kendall's rank-order correlation coefficient, and root-mean-square error.
- Open Access
- 5 September 2020
- Published by Technoscience Academy in International Journal of Scientific Research in Computer Science, Engineering and Information Technology
- p. 69-76
Identification and Recognition of facial images
Face detection is the power to identify a face and recognition is the ability to recognize whose face it is by means of facial characteristics. Face is multivariate and requires a lot of mathematical summation. Almost all imperative applications use a face recognition system. There are many methods that have been already proposed which provides low recognition rate. Hence, the main task of research is to develop a face recognition system with higher recognition capability and better accuracy. This paper proposes Face recognition system by combining two techniques Viola Jones and Principal Component Analysis. An approach of Eigen faces is employed in Principle Component Analysis(PCA). The face recognition system is implemented in MATLAB.
- 5 September 2020
- Published by Technoscience Academy in International Journal of Scientific Research in Computer Science, Engineering and Information Technology
- p. 69-76
Open AccessIdentification and Recognition of facial images
Face detection is the power to identify a face and recognition is the ability to recognize whose face it is by means of facial characteristics. Face is multivariate and requires a lot of mathematical summation. Almost all imperative applications use a face recognition system. There are many methods that have been already proposed which provides low recognition rate. Hence, the main task of research is to develop a face recognition system with higher recognition capability and better accuracy. This paper proposes Face recognition system by combining two techniques Viola Jones and Principal Component Analysis. An approach of Eigen faces is employed in Principle Component Analysis(PCA). The face recognition system is implemented in MATLAB.
- 20 June 2020
- Published by Springer Nature in Multimedia Tools and Applications
- Vol. 79 (33-34) , 24393-24412
Image complexity analysis with scanpath identification using remote gaze estimation model
Analysis of gaze points has been a vital tool for understanding varied human behavioral pattern and underlying psychological processing. Gaze points are analyzed generally in terms of two events of fixations and saccades that are collectively termed as scanpath. Scanpath could potentially establish correlation between visual scenery and human cognitive tendencies. Scanpath has been analyzed for different domains that include visual perception, usability, memory, visual search or low level attributes like color, illumination and edges in an image. Visual search is one prominent area that examines scanpath of subjects while a target object is searched in a given set of images. Visual search explores behavioral tendencies of subjects with respect to image complexity. Complexity of an image is governed by spatial, frequency and color information present in the image. Scanpath based image complexity analysis determines human visual behavior that could lead to development of interactive and intelligent systems. There are several sophisticated eye tracking devices and associated algorithms for recording and classification of scanpath. However, in the present scenario when the chances of viral infections (COVID-19) from known and unknown sources are high, it is very important that the contact less methods and models be designed. In addition, even though the devices acquire and process eye movement data with fair accuracy but are intrusive and costly. The objective of current research work is to establish the complexity of the given set of images while target objects are searched and to present analysis of gaze search pattern. To achieve these objectives a remote gaze estimation and analysis model has been proposed for scanpath identification and analysis. The model is an alternate option for gaze point tracking and scanpath analysis that is non intrusive and low cost. The gaze points are tracked remotely as against sophisticated wearable eye tracking devices available in the market. The model employs easily available softwares and hardware devices. In the current work, complexity is derived on the basis of analysis of fixation and saccade gaze points. Based on the results generated by the proposed model, influence on subjects due to external stimuli is studied. The set of images chosen, act as external stimuli for the subjects during visual search. In order to statistically analyze scanpath for different subjects, certain scanpath parameters have been identified. The model maps and classifies eye movement gaze points into fixations and saccades and generates data for identified parameters. For eye detection and subsequent iris detection voila jones and circular hough transform (CHT) algorithms have been used. Identification by dispersion threshold (I-DT) is implemented for scanpath identification. The algorithms are customized for better iris and scanpath detection. Algorithms are developed for gaze screen mapping and classification of fixations and saccades. The experimentation has been carried on different subjects. Variations during visual search have been observed and analyzed. The present model requires no contact of human subject with any equipment including eye tracking devices, screen or computing devices.
- 20 June 2020
- Published by Springer Nature in Multimedia Tools and Applications
- Vol. 79 (33-34) , 24393-24412
Image complexity analysis with scanpath identification using remote gaze estimation model
Analysis of gaze points has been a vital tool for understanding varied human behavioral pattern and underlying psychological processing. Gaze points are analyzed generally in terms of two events of fixations and saccades that are collectively termed as scanpath. Scanpath could potentially establish correlation between visual scenery and human cognitive tendencies. Scanpath has been analyzed for different domains that include visual perception, usability, memory, visual search or low level attributes like color, illumination and edges in an image. Visual search is one prominent area that examines scanpath of subjects while a target object is searched in a given set of images. Visual search explores behavioral tendencies of subjects with respect to image complexity. Complexity of an image is governed by spatial, frequency and color information present in the image. Scanpath based image complexity analysis determines human visual behavior that could lead to development of interactive and intelligent systems. There are several sophisticated eye tracking devices and associated algorithms for recording and classification of scanpath. However, in the present scenario when the chances of viral infections (COVID-19) from known and unknown sources are high, it is very important that the contact less methods and models be designed. In addition, even though the devices acquire and process eye movement data with fair accuracy but are intrusive and costly. The objective of current research work is to establish the complexity of the given set of images while target objects are searched and to present analysis of gaze search pattern. To achieve these objectives a remote gaze estimation and analysis model has been proposed for scanpath identification and analysis. The model is an alternate option for gaze point tracking and scanpath analysis that is non intrusive and low cost. The gaze points are tracked remotely as against sophisticated wearable eye tracking devices available in the market. The model employs easily available softwares and hardware devices. In the current work, complexity is derived on the basis of analysis of fixation and saccade gaze points. Based on the results generated by the proposed model, influence on subjects due to external stimuli is studied. The set of images chosen, act as external stimuli for the subjects during visual search. In order to statistically analyze scanpath for different subjects, certain scanpath parameters have been identified. The model maps and classifies eye movement gaze points into fixations and saccades and generates data for identified parameters. For eye detection and subsequent iris detection voila jones and circular hough transform (CHT) algorithms have been used. Identification by dispersion threshold (I-DT) is implemented for scanpath identification. The algorithms are customized for better iris and scanpath detection. Algorithms are developed for gaze screen mapping and classification of fixations and saccades. The experimentation has been carried on different subjects. Variations during visual search have been observed and analyzed. The present model requires no contact of human subject with any equipment including eye tracking devices, screen or computing devices.