Abstract

The accuracy of facial emotion recognition systems largely depends on the quality of input images, particularly in interactive customer services where real-time data capture is generally noisy, low in lighting, and resolution-constrained. This paper proposes FIRASS (Facial Image Refinement and Artifact Suppression System) as an integrated pre-processing system that can mitigate these problems by multi-stage facial image denoising and enhancement. FIRASS combines median-Gaussian hybrid denoising, contrast enhancement by CLAHE adaptation, illumination adjustment by Retinex theory, and edge sharpening by unsharp masking. These stages work in concert to slow down visual artifacts while preserving expressive facial features critical for the detection of emotions. Comparative research using typical filters such as median, Gabor, Wiener, and Laplacian indicates that FIRASS produces cleaner and structurally closer-to-reality facial representations. As a result, emotion detection models are provided with input of better quality, improving identification accuracy as well as reliability in real-world customer-facing applications.

Authors

G. Kalaivani, K. Krishnaveni
Sri S. Ramasamy Naidu Memorial College, India

Keywords

Facial Image-Denoising, Emotion-Recognition, FIRASS, PreProcessing, Artifact-Suppression, Customer Service AI, Contrast Enhancement, Illumination-Correction, Facial-Expression Analysis