A Hybrid Framework for Efficient Image Compression Using Autoencoder Integrated with NK-RLE and Clustering Technique

Abstract

Image compression plays a pivotal role in reducing storage requirements and optimizing bandwidth usage without significantly compromising visual quality. This research presents a novel approach to image compression by combining deep learning techniques, clustering methods, and lossless encoding algorithms. The proposed framework utilizes an autoencoder to generate a latent space representation of the image, reducing its dimensionality while preserving essential features. K-Means clustering is employed to group similar features, enhancing compression efficiency, followed by NK-RLE (Non-Keyed Run-Length Encoding) for further lossless data compression. The decoding process reconstructs the compressed image using cluster-based reconstruction and the autoencoder's decoder. The algorithm's performance is evaluated by comparing the original and reconstructed images using Peak Signal-to-Noise Ratio (PSNR) and Mean Squared Error (MSE) metrics. MATLAB software is utilized to simulate and validate the process, providing insights into the effectiveness of combining deep learning and clustering methods for robust image compression. The proposed approach demonstrates potential for practical applications in storage-efficient and bandwidth-optimized image transmission.