The 3rdInternational Conference on Electromechanical Engineering ICEE’2018, November 21-22, 2018,Skikda Article Topic: Robotics and Control
1
Image compression of Surface defects of the hot-rolled steel strip using discrete Cosine Transform and Discrete Wavelet
Transform
Adel Boudiaf
1, a*, Kadri Oussama
2, b,Rachid Zaghdoudi
1,c, SlimaneZiani
1, d, Dehimi Said
1, e1Research Center in Industrial Technologies CRTI P.O.Box 64, cheraga 16014 Algiers, Algeria
2Electrical engineering department, Mohamed Khider University, 07000 Biskra,Algeria
aa.boudiaf@crti.dz / csc.boudiaf@gmail.com
bkadrioussama86@gmail.com
cr.zaghdoudi@crti.dz
ds.ziani@crti.dz
ABSTRACT—Le contrôle automatique de la qualité de la surface d’acier laminé à chaud à l’aide de systèmes de vision par ordinateur est une application en temps réel,ce qui nécessite des techniques de compression d’images très efficaces afin d’améliorer la capacité de stockage et de transmission des données. Dans ce travail, deux méthodes de compression d’image sont simulées. Il s'agit de la transformation en cosinus discrète (DCT) et de la transformation en ondelettes discrète (DWT). Les résultats de la simulation montrent que la technique de transformation en ondelettes discrètes (DWT) donne un résultat amélioré par rapport à la transformation en cosinus discrète (DCT).
MOTS-CLÉS : Control automatique de qualité, Vision par ordinateur, compression d’image, transformée en cosinus, transformée en ondelettes.
ABSTRACT. Automatic quality control of surface of hot rolled steel using computer vision systems is a real time application, which requires highly efficient Image compression techniques in order to improve the data transmission and storage capacity.In this work, two image compression methods are simulated. They are Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT).The results of simulation are shown that The Discrete Wavelet Transform (DWT) technique given improved result compared with Discrete Cosine Transform (DCT).
KEYWORDS:Automatic quality control, Computer vision, image compression, Cosine transform, wavelets transform.
1. Introduction
Image monitoring of rolling mill products is a fundamental axis of development and industrial research. Its purpose is to give knowledge about the condition of products of rolling mill at each moment without stopping the production line. The monitoring allows avoiding the production losses related to surface defects and improved the production and productivity [1-3]. However, these images are often volumetric, which is poses a problem in quality control centers, where engineers are unable to store these large quantities of images, so they need efficient compression techniques for archiving and transmitting them.
.Image compression is divided into two types: Lossy compression and Lossless compression [4-5].Lossless methods aim to create a compressed image strictly identical to the original without losing any of its features. It is mostly used in medical imaging, storage of legal records and ZIP file format. This type of compression can be obtained by the application of different algorithms: Run Length coding (RLC), Huffman coding, Lempel Ziv Welch coding (LZW), among others. However, it has been revealed that those techniques exhibit some limitations. For instance, Run Length Coding (RLC) is not suitable for continuous-tone Images such as photographs. Likewise, the effectiveness of Huffman coding depends upon the precision of the statistical model employed. This makes Huffman-based compression a rather relatively slow process [5-7]. Likewise, the effectiveness of Huffman coding depends upon the precision of the statistical model employed. This makes Huffman-based compression a rather relatively slow process [5-7]. In lossy compression, the compressed image is not a replica of the original image. Indeed, certain statistical details of the latter can be discarded without ruining the original image appearance. This type of compression is mostly used in broadcast
