Abstract—Current advances in research are the direction of new alternatives, combining different methods to communicate quickly and store a growing amount of data more efficiently: this mostly comprises the digital images which have invad our daily lives. That is why, this paper is devoted to the presentation of a digital compression algorithm based on improving the exploitation of the advantages of existing algorithms. It also aims to contribute to improving the performance of the transmission, and ensuring storage with a high compression ratio and low distortion.
In an initial stage, as pre-treatment, the image is decomposed into sub-bands (approximation sub-band + various details sub-bands), using a wavelet transform biorthogonal. Then, comes the phase of compression: Discrete Cosine Transform (DCT) is used to encode the approximation sub-band, which contains most of the information, followed by scalar quantization. In the last stage, the details of the sub-bands are encoded using vector quantization grouped into sub-bands using neural competitive learning.
I want to end this paper by demonstrating the effectiveness of this method, which is applied to compression of voluminous images, remote sensing, provided by the station of the National Meteorology Office in the visible, and infrared channels covering the area (+ 47˚ 19˚ North and South and West 16˚East 16˚). For compression ratio values of over 97.2994%, experimental results obtained show that these are faithfully reproduced, and these are not accompanied by artifact. Moreover, the storage memory has been significantly reduced.
Index Terms—Image compression, wavelet, DCT, competitive neuronal network, vector quantization, coding.
Amel Bey Boumezrag is with the Department of Electronic Engineering, Faculty of Science, University of Laghouat Algeria, BP 37G, Algeria (e-mail: am_by@yahoo.fr).
[PDF]
Cite:Amel Bey Boumezrag, "An Improved Algorithm for Image Compression Based on Biorthogonal Wavelet, DCT and Competitive Neuronal Networks," International Journal of Computer Theory and Engineering vol. 9, no. 2, pp. 87-91, 2017.