ICTACT Journal on Image and Video Processing

M. Velammal ¹, Thiyam Ibungomacha Singh ², Nilesh Madhukar Patil ³, Subharun Pal ⁴

Affiliations
1 Department of Computer science and Technology, Karpagam College of Engineering, IN
2 Department of Computer Science and Engineering, Manipur Institute of Technology, Manipur University Campus, IN
3 Department of Computer Engineering, Dwarkadas Jivanlal Sanghvi College of Engineering, IN
4 Department of Computer Science and Engineering, Indian Institute of Technology Jammu, IN

Abstract

This paper introduces a novel approach for multiframe image restoration using Generative Adversarial Networks (GANs). Traditional image restoration techniques often struggle with handling complex degradation patterns and noise in images. In contrast, GANs have demonstrated remarkable capability in generating realistic and high-quality images. The proposed method leverages the power of GANs to restore multiframe degraded images by training the generator to learn the underlying clean image from a set of degraded frames. The discriminator collaborates with the generator to ensure the fidelity of the restored output. Experimental results on various datasets show that the proposed multiframe image restoration approach achieves superior performance compared to state-of-the-art methods in terms of image quality and fidelity.

Keywords

Multiframe, Image Restoration, Generative Adversarial Networks (GANs), Degradation Patterns, Fidelity.

Full Text

     

N. Ananthi ¹, Thiyam Ibungomacha Singh ², Nihar Ranjan Behera ³, R. K. Gnanamurthy ⁴

Affiliations
1 Department of Information Technology, Easwari Engineering College, IN
2 Department of Computer Science and Engineering, Manipur Institute of Technology, IN
3 Geneva Business Center, Swiss School of Business and Management, CH
4 Department of Electronics and Communication Engineering, VSB College of Engineering Technical Campus, IN

Abstract

This research presents a novel approach that combines the C4.5 algorithm with Adversarial Learning-based Adaptive Data Augmentation (ADA) for Color and Multispectral Processing, leading to a significant enhancement in Image Analysis. The C4.5 algorithm, known for its decision tree construction, is integrated with ADA, which employs adversarial learning principles to generate diverse and realistic training samples. This integration enables the augmentation of both color and multispectral images, effectively boosting the robustness and accuracy of image analysis tasks. The proposed method showcases improved performance in various applications such as object recognition, classification, and scene understanding. Experimental results demonstrate the superiority of the proposed approach compared to traditional methods, substantiating its potential for advancing image analysis techniques.

Keywords

C4.5 algorithm, Adversarial Learning, Adaptive Data Augmentation (ADA), Color Processing, Multispectral Processing, Image Analysis

Full Text