Asian Journal of Research in Computer Science

Handicraft Bell metal products hold great cultural and artistic importance in Assamese society, especially in the Sarthebari region, where they are crafted using traditional techniques passed down through generations.However, studying and classifying images of these intricate products using modern machine learning methods comes with challenges. Variations in pixel intensity, caused by changes in brightness and color during photography, can lower image quality. Additionally, the detailed textures and complex backgrounds of these products make it difficult for computers to separate the main object from its surroundings. A dataset of fifty handicraft bell metal product images was collected directly from production units in Sarthebari area using digital camera. Image pre-processing is essential to increase the model performance. This study examines five contrast enhancement techniques-Histogram Equalization (HE), Adaptive Histogram Equalization (AHE), Unsharp Masking (UM), Contrast Limited Adaptive Histogram Equalization (CLAHE), and Triangular Fuzzy Membership Contrast Limited Adaptive Histogram Equalization (TFM-CLAHE). The performance of these techniques was evaluated using four quantitative metrics: Mean Squared Error (MSE), Signal-to-Noise Ratio (SNR), Peak Signal-to-Noise Ratio (PSNR), and Similarity Index (SI) . The results show that UM and TFM-CLAHE are the most effective methods for enhancing image details while maintaining clarity. These techniques help to highlight the intricate designs of bell metal products, making them useful for automated classification and quality control. By applying these methods, technology can better support the preservation and promotion of this ancient Assamese craft.