Surface defect detection is crucial for ensuring high quality and minimizing defects in industrial products. In this paper, we propose a lightweight hybrid approach for detecting surface defects in industrial RAW images, optimized for quality control. The method integrates traditional image processing techniques, such as contrast enhancement, differential imaging, and morphological operations, with a machine learning model combining Histogram of Oriented Gradients (HOG) for feature extraction and Support Vector Machines (SVM) for classification. Experimental evaluations on real-world datasets show an accuracy of 96.08%, demonstrating robustness and practical effectiveness. The system operates efficiently using only CPU resources, with an average processing time of 80 ms per image on an Intel Core i5 CPU (8 GB RAM), without requiring GPU acceleration. Compared to deep learning methods, it significantly reduces computational costs and is suitable for real-time, cost-effective industrial quality control on platforms like Jetson Nano and Raspberry Pi.