Optimal deep learning based image compression technique for data transmission on industrial Internet of things applications
Ben Sujitha,Velmurugan Subbiah Parvathy,E. Laxmi Lydia,Poonam Rani,Zdzislaw Polkowski,K. Shankar
DOI: https://doi.org/10.1002/ett.3976
IF: 3.6
2020-04-14
Transactions on Emerging Telecommunications Technologies
Abstract:<p>Recently, industrial Internet of things becomes more popular and it involves a group of intelligent devices linked to create systems which observe, gather, communicate, and investigate data. In this view, the demand for compression techniques in remote sensing images is increasing since low complexity technique is required in spacecraft. Deep learning, for instance, convolutional neural network (CNN) has gained more attention in the domain of computer vision, particularly for high‐level applications like detection along with interpretation. At the same time, it is difficult to resolve the low‐level applications like image compression and it is investigated in this article. This article presents an optimal compression technique using CNNs for remote sensing images. The proposed method uses CNN for learning the compact representation of the original image which held the structural data and was then coded by Lempel Ziv Markov chain algorithm. Next, the encoded image was reconstructed to retrieve the original image with high reconstructed image quality. The proposed optimal compression technique is compatible with the available image codec standards. Wide range of experiments was carried out and the results were compared with binary tree and optimized truncation, JPEG, and JPEG2000 in terms of compression efficiency, reconstructed image quality, and space saving (SS). The obtained results apparently proved the effectiveness of the presented method, which attains an average peak signal to noise ratio of 49.90 dB and SS of 89.38%.</p>
telecommunications