Optimal deep learning based image compression technique for data transmission on industrial Internet of things applications

Abstract

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%.

Solving Image Processing Critical Problems Using Machine Learning

“…While applying neural network method in image we can reconstruct and compress the image from the following provided image data set on which machine learning model is trained [60]. Fast data transmission in industrial scale on IoT devices is possible due to machine learning algorithm for the image compression [61]. In medical science for the detection of diabetic retinopathy is possible because of the image compression and machine learning image processing plays a vital role in the development of better treatment for diabetic patients [62].…”

Comprehensive review of hyperspectral image compression algorithms

“…CNN-NTD 31 algorithm discussed in tensor decomposition method can also be categorized under learning-based compression due to the use of deep learning (CNN) in its first step. An application-oriented compression has been proposed by Sujitha et al 88 that used the concepts of Lempel Ziv Markov chain algorithm (LZMA) coder to generate the bitstreams. CNN-LZMA is the algorithm that learns to generate the compact representation of the raw 3-D image.…”

DBTP2SF: A deep blockchain‐based trustworthy privacy‐preserving secured framework in industrial internet of things systems

“…The first level introduces enhanced proof‐of‐work (ePoW) algorithm of BC services, to perform data authentication and prevention from poisoning attacks. The second level uses an AutoEncoder (AE) technique to transform features into encoded format for mitigating inference attack that can learn from DL 35,36 . This two‐level technique improves the overall performance of privacy‐preserving approach and is used as an input of deep neural networks (DNNs) for classifications of normal and attack vectors 37 .…”