Year: 2025
Type: article
Abstract: In many image processing applications, Blind Image Quality Assessment (BIQA) plays a crucial role when the reference image is unavailable. However, existing BIQA methods often lack consistency with human perception and are limited to specific types of distortion. This research introduces a novel approach for image quality assessment by combining Convolutional Neural Network (CNN) feature extraction and Extreme Learning Machine (ELM) regression. By leveraging a pre-trained ResNet CNN model, features are extracted from distorted images and used to train the ELM model. The proposed method achieves accelerated convergence in deep model training and significantly enhances accuracy in predicting image quality compared to traditional regression methods. Experimental results demonstrate the superiority of the proposed approach over state-of-the-art BIQA methods, making it a promising solution for automated image quality evaluation in various applications such as image processing and computer vision.
Institutions: Universiti Malaysia Perlis, Thaksin University
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Funder: Universiti Malaysia Perlis
OpenAlex ID: W4409609091