Enhancing Image Quality Using a Blind Quality Evaluator Custom Network for Object Detection

Image quality enhancement is a critical task in computer vision and image processing, particularly in poor-lighting conditions where noise and distortion often degrade the perceptual quality. This paper presents a novel approach for image quality enhancement using a cosine quality evaluator custom network for indoor object detection. Adaptive filtering techniques, such as Cosine Adaptive Enhanced Block-Matching and 3D filtering, focus on denoising to effectively reduce noise while preserving key image features. The Blind/Reference Image Spatial Quality Evaluator score allows for effective evaluation without a reference image by examining statistical qualities and distortion characteristics to offer a no-reference measure for evaluating image quality. Deep custom convolutional neural networks are employed after denoising, classifying images based on their enhancement score to enhance the quality. This process ensures that important details, such as edges and textures, are retained while improving overall image clarity. Despite the difficulties caused by high computational costs, real-time processing is achieved through efficient denoising and quality assessment methods. The combination of these methods allows for effective image enhancement, generally in complex situations involving blur, noise, and low-light conditions, with improved processing efficiency and perceptual quality. At 99.53% accuracy and a 99.62% F1-score, the proposed Blind Quality Evaluator Custom Network (BQECN) approach achieves exceptional results. The model shows robustness with an accuracy of 99.86% and an F1-score of 99.76% in 10-fold cross-validation. These outcomes demonstrate the method’s ability to handle blurred and noisy images.