Title: Improved Indian currency recognition: neighbourhood-centred image processing and CNNs with region of pixel selection techniques

Authors: Bhargavi Mopuru; Anurag Sinha; Ankita Agrawal; Amritesh Nandan; Ankit Agarwal; Peddi Nikitha; Vandana Sharma; Ankit Anand; Doğan Keskin; Biresh Kumar; Pooja Jha

Addresses: Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation, Andhra Pradesh, 520002, India ' Department of Computer Science, IGNOU, India ' Department of Commerce, Indian Institute of Management Kozhikode Kerala, Kozhikode, India ' Department of Computer Science, Graphic Era (Deemed to be University), Uttarakhand, 248002, India ' Department of Computer Science and Engineering, Kalasalingam Academy of Research and Education, Krishnankoil, Virudhunagar, Tamil Nadu, India ' Department of Computer Science and Engineering, Kalasalingam Academy of Research and Education, Krishnankoil, Virudhunagar, Tamil Nadu, India ' Department of Computer Science, CHRIST (Deemed to be University), Delhi NCR, India ' Department of Computer Science, Amity University, Jharkhand, India ' Department of Computer Science, Turkish German University, Istanbul, Turkey ' Department of Information Technology, Amity University Jharkhand, Ranchi, India ' Department of Information Technology, Amity University Jharkhand, Ranchi, India

Abstract: The paper proposes an improved approach for Indian currency recognition using neighbourhood-centred image processing and convolutional neural networks (CNNs) with region of pixel selection techniques. The method includes image pre-processing steps such as noise reduction, contrast enhancement, and resizing. A neighbourhood-centred image processing technique is applied to capture contextual information from local neighbourhoods around each pixel. A CNN-based model is then trained on the pre-processed images to learn discriminative features for currency recognition. To enhance accuracy and efficiency, a region of pixel selection technique is introduced to select only relevant regions of interest for CNN training and inference, reducing computational overhead. Experimental results demonstrate the effectiveness of the proposed approach, achieving high accuracy in currency recognition and improved efficiency in terms of computational time and memory requirements. The proposed method has potential applications in automated cash-handling machines, vending machines, and mobile payment systems where reliable currency recognition is essential.

Keywords: currency recognition; image processing; RGB; threshold; decision boundary; CNN; image processing; features.

DOI: 10.1504/IJSEM.2025.148476

International Journal of Services, Economics and Management, 2025 Vol.16 No.4/5, pp.482 - 516

Received: 17 Apr 2023
Accepted: 25 Sep 2023
Published online: 08 Sep 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article