Ijaz Ahmad (Ph.D.)

Ijaz Ahmad (Ph.D.)

Postdoctoral Researcher @ Korea Univ.

Sitemap

A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.

Pages

Posts

Future Blog Post

less than 1 minute read

Published:

This post will show up by default. To disable scheduling of future posts, edit config.yml and set future: false.

Blog Post number 4

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 3

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 2

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 1

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

portfolio

publications

Color-to-Grayscale Algorithms Effect on Edge Detection—A Comparative Study

Published in IEEE International Conference on Electronics, Information, and Communication (ICEIC), 2018

Abstract

In image processing, color images are converted into grayscale to perform edge detection, without considering the color-to-grayscale algorithms in details. We have evaluated the impact of various color-to-grayscale algorithms in edge detection. This study shows that edges are not only dependent on the methods used for edge detection but also on the color-to-grayscale conversion algorithms. We have implemented ten different color-to-grayscale conversion algorithms in MATLABR2016a and the resultant grayscale images were further tested with eight different edge detection algorithms. The experimental results shows that the Lightness color-to-grayscale conversion algorithm achieves higher performance among evaluated methods.

Just-Noticeable-Difference Based Edge Map Quality Measure

Published in KINGPC International Conference on Next Generation Computing (ICNGC), 2018

Abstract

The performance of an edge detector can be improved when assisted with an effective edge map quality measure. Several evaluation methods have been proposed resulting in different performance score for the same candidate edge map. However, an effective measure is the one that can be automated and which correlates with human judgement perceived quality of the edge map. Distance-based edge map measures are widely used for assessment of edge map quality. These methods consider distance and statistical properties of edge pixels to estimate a performance score. The existing methods can be automated; however, they lack perceptual features. This paper presents edge map quality measure based on Just-Noticeable-Difference (JND) feature of human visual system, to compensate the shortcomings of distance-based edge measures. For this purpose, we have designed constant stimulus experiment to measure the JND value for two spatial alternative. Experimental results show that JND based distance calculation outperforms existing distance-based measures according to subjective evaluation.

Region-based Selective Compression and Selective Encryption of Medical Images

Published in ACM International Conference on Smart Media and Applications (SMA), 2020

Abstract

Image compression and encryption1are two processes that enable telemedicine application of eHealth services. However, performing these operations on the whole content of an image is computationally expensive. This work proposes a method for selective compression and selective encryption of medical images. It is based on lossless compression and encryption of the region of interest (ROI) in medical images. The non-ROI part of the image is compressed in lossy mode and is stored or transmitted as plain data, in order to further reduce the image size and to avoid the computational cost of encrypting huge volumes of medical images. Our analysis shows that the proposed method provides the necessary security and is secured against various attacks. In addition, the compression savings achieved by the proposed method is about 28% while preserving the crucial information in the ROI for correct diagnosis. For a quality factor of 80%, the reconstructed image has a peak signal-to-noise ratio of 42.5 dB. The proposed method requires less computational resources and enables processing of huge volume of image data in low power network.

An Approach to Run Pre-Trained Deep Learning Models on Grayscale Images

Published in IEEE International Conference on Artificial Intelligence in Information and Communication (ICAIIC), 2021

Abstract

Transfer learning helps the performance of a learning algorithm significantly when training deep learning models on challenging datasets. However, the pre-trained networks have certain constraints in terms of their architecture. For example, due to the wide availability of color images, stateof- the-art pre-trained networks expect an input image with three color channels. Grayscale images have small sizes as compared to color images and thus can enable real time computer vision applications in scenarios where there are constraints on device memory and bandwidth. Therefore, in this work we propose an approach to run pre-trained models on grayscale images for image classification tasks. We have used the VGG16 pre-trained model to classify Kaggle Dogs vs. Cats dataset. We have compared our results with VGG16 applied on color images. Our results have shown that when the weights for the first hidden layer are initialized as the mean of the pretrained network weights then the classification accuracy with only 0.04% error can be achieved. Our analysis has shown that comparable benefits can be reaped when using grayscale images for deep learning based classification tasks with only one-third of the bandwidth and storage requirements.

A novel hybrid image encryption–compression scheme by combining chaos theory and number theory

Published in Signal Processing: Image Communication, 2021

Abstract

Compression and encryption are often performed together for image sharing and/or storage. The order in which the two operations are carried out affects the overall efficiency of digital image services. For example, the encrypted data has less or no compressibility. On the other hand, it is challenging to ensure reasonable security without downgrading the compression performance. Therefore, incorporating one requirement into another is an interesting approach. In this study, we propose a novel hybrid image encryption and compression scheme that allows compression in the encryption domain. The encryption is based on Chaos theory and is carried out in two steps, i.e., permutation and substitution. The lossless compression is performed on the shuffled image and then the compressed bitstream is grouped into 8-bit elements for substitution stage. The lossless nature of the proposed method makes it suitable for medical image compression and encryption applications. The experimental results shows that the proposed method achieves the necessary level of security and preserves the compression efficiency of a lossless algorithm. In addition, to improve the performance of the entropy encoder of the compression algorithm, we propose a data-to-symbol mapping method based on number theory to represent adjacent pixel values as a block. With such representation, the compression saving is improved on average from 5.76% to 15.45% for UCID dataset.

Fine-Tuning Pre-Trained Deep Learning Models for Multiclass Grayscale Images Classification

Published in KINGPC International Conference on Next Generation Computing (ICNGC), 2021

Abstract

Transfer learning significantly improves the performance of a deep learning model on challenging datasets. However, the pre-trained models have certain constraints in terms of their architecture. For example, the state-of-the-art pre-trained models expect an input image with three-color channels because of the wide availability of color images. However, there are certain domains, e.g., medical applications, where grayscale images are produced and the models are required to perform certain tasks on them. Therefore, in this work we propose an approach to run pre-trained models on grayscale images while benefiting from transfer learning for multiclass classification task. We have used the MobileNetV2 pre-trained model to classify the CIFAR datasets. We have compared our results with a conventional method where the grayscale image is stacked up to form a pseudo-color image. Our analysis have shown that the proposed method reduces the computational time per epoch while improves the accuracy of the model.

Encryption-then-Compression System for Cloud-based Medical Image Services

Published in IEEE International Conference on Information Networking (ICOIN), 2022

Abstract

In the recent years, encryption-then-compression (EtC) algorithms are becoming popular for multimedia applications due to their low computational requirements and format compliancy with the JPEG image standard. In this study, we first extend the applications of EtC algorithms to medical image processing domain. Then, given importance of medical data, we propose a method to improve the security performance of the conventional EtC schemes with acceptable degradation in the compression savings but without compromising quality of the recovered images. We have implemented a deep learning (DL) model for tuberculosis (TB) screening in Montgomery datasets. Our analysis shows that the recovered images are of high quality and has no adverse effect on the accuracy of the DL model for TB detection. The proposed method preserves the crucial information in the image necessary for correct diagnosis while significantly reducing the image size and providing security to enable telemedicine application of eHealth services.

Determining Jigsaw Puzzle State from an Image based on Deep Learning

Published in IEEE International Conference on Artificial Intelligence in Information and Communication (ICAIIC), 2022

Abstract

The intriguing nature of jigsaw puzzle has captured the attention of researchers for many years. In this paper, we propose a deep learning model to determine different states of jigsaw puzzle from an image. We represent the task as a classification problem where each state of the puzzle is considered as a class. For this purpose, we have proposed a method to generate a dataset that can efficiently represent the jigsaw puzzle states space. The proposed model has 93% accuracy on the test dataset. In addition, we have shown that when the tiles size changes the model is still able to recognize 83% of the states. Though, genetic algorithms (GA) have been successful in solving larger puzzles, they require hand-crafted sophisticated compatibility scores. The computation and memory requirement to store the piecewise compatibility measure increases with the size of the puzzle. As an application, we have shown that the proposed method can be used as a fitness function of GA based jigsaw puzzle solver without using any compatibility measure.

Privacy-Preserving Surveillance for Smart Cities

Published in IEEE International Confernce on Ubiquitous and Future Networks (ICUFN), 2022

Abstract

This work presents privacy-preserving surveillance system for smart cities based on perceptual encryption algorithm (PE). The encryption is block-based and provides necessary level of security while preserving intrinsic properties of an image necessary for compression. Unlike existing PE methods, the proposed method retains color information, thus can enable processing in encryption domain. The analysis shows that our method achieves same compression performance as existing methods while providing better security. In addition, we have performed face recognition on the encrypted images and demonstrated that the proposed method delivers the same recognition accuracy as that of the plain images.

A Perceptual Encryption-Based Image Communication System for Deep Learning-Based Tuberculosis Diagnosis Using Healthcare Cloud Services

Published in Electronics, 2022

Abstract

Block-based perceptual encryption (PE) algorithms are becoming popular for multimedia data protection because of their low computational demands and format-compliancy with the JPEG standard. In conventional methods, a colored image as an input is a prerequisite to enable smaller block size for better security. However, in domains such as medical image processing, unavailability of color images makes PE methods inadequate for their secure transmission and storage. Therefore, this study proposes a PE method that is applicable for both color and grayscale images. In the proposed method, efficiency is achieved by considering smaller block size in encryption steps that have negligible effect on the compressibility of an image. The analyses have shown that the proposed system offers better security with only 12% more bitrate requirement as opposed to 113% in conventional methods. As an application of the proposed method, we have considered a smart hospital that avails healthcare cloud services to outsource their deep learning (DL) computations and storage needs. The EfficientNetV2-based model is implemented for automatic tuberculosis (TB) diagnosis in chest X-ray images. In addition, we have proposed noise-based data augmentation method to address data deficiency in medical image analysis. As a result, the model accuracy was improved by 10%.

Perceptual Encryption-based Privacy-Preserving Deep Learning in Internet of Things Applications

Published in IEEE International Conference on Information and Communication Technology Convergence (ICTC), 2022

Abstract

In the Internet of Things (IoT) ecosystem, cloud computations are widely utilized to train a deep learning model on public datasets. The trained model can then be deployed on edge devices as an inference engine to facilitate various IoT applications in real-time. For consistent accurate predictions, the model needs to be retrained on the most recent data. This necessitates the data share between cloud servers and edge devices. However, the data acquired by IoT end devices usually consists of sensitive information and sharing them with cloud services provider results in users' privacy issues. In addition, the exchange of large volume of data requires high bandwidth. This study proposes an extension of block-based perceptual encryption (PE) algorithm to enable DL computations in encryption domain in order to protect users’ privacy. For this purpose, four different extensions of baseline PE methods were analyzed in terms of compression savings, encryption efficiency and privacy-preserving DL performance. The analyses have shown that when global contents are preserved in each color channel, the extended PE method offers better security and preserves compression savings with only 3% drop in accuracy as compared to baseline method.

Performance Analysis of Cloud-based Deep Learning Models on Images Recovered without Channel Correction in OFDM System

Published in IEEE Asia Pacific Conference on Communication (APCC), 2022

Abstract

Channel correction plays an important role in performance of wireless communication systems. In conventional systems, channel estimation is one of the blocks at receiver side to compute channel impulse response. Various algorithms have been proposed to make them efficient and improve their performance. However, precise channel estimation incurs additional computational cost and increases complexity of the overall system. In this study, we have considered the otherwise to bypass channel estimation of an Orthogonal Frequency Division Multiplexing (OFDM) based image communication system designed to enable cloud-based deep learning (DL) computation. The simulations present performance analysis of OFDM system with and without channel correction in-terms of bit error rate (BER), and two image quality measures. Recovered image quality difference between the two systems significantly increases with higher Eb/N0. For inferencing analysis, in higher Eb/N0 regions, model performance on images recovered with correction is same as on the original images while lags behind by 6% on images without correction. In lower Eb/N0 regions, the model accuracy reduces by 10% on average for both systems. In addition, the model accuracy shows overlapping pattern in that region and for 3 dB, it has performed better on images recovered without correction.

IIB–CPE: Inter and Intra Block Processing-Based Compressible Perceptual Encryption Method for Privacy-Preserving Deep Learning

Published in Sensors, 2022

Abstract

Perceptual encryption (PE) of images protects visual information while retaining the intrinsic properties necessary to enable computation in the encryption domain. Block–based PE produces JPEG-compliant images with almost the same compression savings as that of the plain images. The methods represent an input color image as a pseudo grayscale image to benefit from a smaller block size. However, such representation degrades image quality and compression savings, and removes color information, which limits their applications. To solve these limitations, we proposed inter and intra block processing for compressible PE methods (IIB–CPE). The method represents an input as a color image and performs block-level inter processing and sub-block-level intra processing on it. The intra block processing results in an inside–out geometric transformation that disrupts the symmetry of an entire block thus achieves visual encryption of local details while preserving the global contents of an image. The intra block-level processing allows the use of a smaller block size, which improves encryption efficiency without compromising compression performance. Our analyses showed that IIB–CPE offers 15% bitrate savings with better image quality than the existing PE methods. In addition, we extended the scope of applications of the proposed IIB–CPE to the privacy-preserving deep learning (PPDL) domain.

Perceptual Encryption-based Privacy-Preserving Deep Learning for Medical Image Analysis

Published in IEEE International Conference on Information Networking (ICOIN), 2023

Abstract

The widespread adoption of deep learning (DL) solutions in the healthcare organizations is obstructed by their compute intensive nature and dependability on massive datasets. In this regard, cloud–services such as cloud storage and computational resources are emerging as an effective solution. However, when the image data are outsourced to avail such services, there is a privacy concern that the data should be kept protected not only during transmission but during computations as well. To meet these requirements, this study proposed a privacy-preserving DL (PPDL) scheme that enable computations without the need of decryption. The encryption is based on perceptual encryption (PE) that only hides the perceivable information in an image while preserves other characteristics that are necessary for DL computations. Precisely, we have implemented a binary classifier based on EfficientNetV2 for the COVID-19 screening in the chest X-ray (CXR) images. For the PE algorithm, the suitability of two pixelbased and two block-based PE methods was analyzed. The analysis have shown that when global contents are left unmodified (pixel-based PE), then the DL-based model achieved the classification accuracy same as that of the plain images. On the other hand, for block-based PE algorithms, there is up to 3% drop in the model’s accuracy and sensitivity scores.

Robustness of Deep Learning enabled IoT Applications Utilizing Higher Order QAM in OFDM Image Communication System

Published in IEEE International Conference on Artificial Intelligence in Information and Communication (ICAIIC), 2023

Abstract

In a standard Orthogonal Frequency Division Multiplex (OFDM)-based communication system, higher order M-ary Quadrature Amplitude Modulation (M-QAM) increases throughput and link capacity significantly due to the larger number of bits per symbol. Such a system is suitable to transmit large volumes of image data for real-time Internet of Things (IoT) applications. However, higher order M-QAM increases the noise margin and Bit Error Rate (BER) in the received signal. Although the error bits corrupt the individual pixels values, the global contents of the image may remain intact and can be interpreted by Deep Learning (DL) models. Therefore, the current study analyzes the robustness of DL models for image communication systems utilizing higher order M-QAM. The performance analysis of the system is presented in terms of the BER and image quality measures. In the analysis, distortions in the reconstructed images decrease significantly when using higher Eb/N0 across all M-QAM. For higher Eb/N0 in the 1024- QAM system, the performance of the model improved by 11% and obtained an accuracy of 86%, lagging by 6% compared to the accuracy obtained on the original images. Overall, the model maintained a high accuracy range between 83% and 90% at Eb/N0 10 dB across all M-QAM.

Deep Learning-based Image Quality Assessment Metric for Quantifying Perceptual Distortions in Transmitted Images

Published in IEEE International Symposium on Communications and Information Technologies (ISCIT), 2023

Abstract

An Image Quality Assessment (IQA) metric measures the quality degradation of an image and is used to optimize parameters of an image processing algorithm. The IQA score can be also an important indicator of the target downstream application performance. With the popularity of deep learning (DL)-based applications in resource constrained domains, most of the DL computations are outsourced to avail remotely located resources. The image data transmission for this purpose is susceptible to distortions due to the imperfect communication environment. The existing IQA metrics used to evaluate the quality of these images mainly rely on human judgment and do not account for the perceptual distortions that are responsible for the degradation in DL model performance. To address this issue, we propose a convolutional autoencoderbased IQA metric that compares images in low dimensional feature space and can be used to monitor image degradation occurred during data transmission. The simulation analysis shows that the proposed method introduces at best 0.13% error while on average 8% error compared to the application model accuracy. Importantly, the proposed IQA score coincides with the DL model performance on a downstream task and can be used to optimize the parameters of a communication system.

Comprehensive Analysis of Compressible Perceptual Encryption Methods—Compression and Encryption Perspectives

Published in Sensors, 2023

Abstract

Perceptual encryption (PE) hides the identifiable information of an image in such a way that its intrinsic characteristics remain intact. This recognizable perceptual quality can be used to enable computation in the encryption domain. A class of PE algorithms based on block-level processing has recently gained popularity for their ability to generate JPEG-compressible cipher images. A tradeoff in these methods, however, is between the security efficiency and compression savings due to the chosen block size. Several methods (such as the processing of each color component independently, image representation, and sub-block-level processing) have been proposed to effectively manage this tradeoff. The current study adapts these assorted practices into a uniform framework to provide a fair comparison of their results. Specifically, their compression quality is investigated under various design parameters, such as the choice of colorspace, image representation, chroma subsampling, quantization tables, and block size. Our analyses have shown that at best the PE methods introduce a decrease of 6% and 3% in the JPEG compression performance with and without chroma subsampling, respectively. Additionally, their encryption quality is quantified in terms of several statistical analyses. The simulation results show that block-based PE methods exhibit several favorable properties for the encryption-then-compression schemes. Nonetheless, to avoid any pitfalls, their principal design should be carefully considered in the context of the applications for which we outlined possible future research directions.

Perceptual Image Encryption: A Communication Perspective

Published in IEEE International Conference on Information Networking (ICOIN), 2024

Abstract

Due to the popularity of cloud services, privacy sensitive image data is outsourced to avail of third-party owned computation and/or storage resources. In this regard, image encryption is necessary not only to protect identifiable information in them that can lead to privacy concerns, but to protect the data ownership as well. Different from the conventional full image encryption techniques, Perceptual Encryption (PE) algorithm protects only perceivable image contents while leaving its intrinsic characteristics intact to cater for the requirements of multimedia applications. Several techniques (such as the pseudo grayscale representation, processing of each color channel independently, and sub-block processing) have been proposed to achieve an efficient tradeoff between security efficiency and the multimedia application’s performance. This study integrates PE algorithms into the source coding block of an orthogonal frequency-division multiplexing (OFDM) system and analyze their performance in terms of the recovered image quality. Specifically, we have considered four PE schemes, and four modulation schemes and two channel models in the OFDM system. Our analyses have shown that PE methods are robust against wireless communication impairments with a slight difference between the quality of recovered images from plain and PE cipher images over a Rayleigh fading channel.

talks

teaching

Teaching experience 1

Undergraduate course, University 1, Department, 2014

This is a description of a teaching experience. You can use markdown like any other post.

Teaching experience 2

Workshop, University 1, Department, 2015

This is a description of a teaching experience. You can use markdown like any other post.