Image Compression and Noise Reduction through Algorithms in Wavelet Domain
Catalin Marian Dumitrescu, Maria Simona Raboacă, Ioana Manta · 2020
Image compression and noise reduction are two important application in the field of digital image processing. The use of transformed domains plays an important role in both mentioned applications, offering a different representation of the data, in which the main characteristics to be processed become more distinct. Image compression aims to reduce the number of bits needed for the digital representation of the image with or without an acceptable loss of image quality. The underlying idea behind the use of linear transforms in signal processing is a more efficient quantization of the coefficients corresponding to the transformed domain, compared to the case of the coefficients corresponding to the primary space. From this point of view, the wavelet transforms allow, due to the multi-resolution representation that they accomplish and the correlation between the sub-images of this representation, a more efficient quantization and entropic encoding of the coefficients corresponding to the transformed domain. In the field of image compression, the multi-resolution analysis property of the wavelet transform allows progressive quality-scalable and spatially scalable transmissions, very useful in applications that require interactive searching or when bandwidth transmission channels are available. The wavelet transform focuses most of the signal energy in the low pass sub-band with the lowest resolution, the coefficients of this sub-band having high absolute values. The rest of the coefficients obtained after decomposition have small absolute values, being canceled after the quantization process. A good compression rate can be obtained in this case if the absolute values of the coefficients and their locations are encoded separately.