Analysis of image compression using singular value decomposition

TImage processing has become a more crucial area nowadays due to the advancement of imaging technology. Although image quality is an essential criterion, in some cases, it leads to a storage bottleneck. Thus, an efficient method is needed to reduce the amount of data while maintaining quality. One preferred method is image compression using Singular Value Decomposition (SVD). In this method, the image matrix is reduced by rank, k. However, the right value of k is still debatable since it is based on the image features. Therefore, this study did some analysis to choose an ideal k. Three popular images-Afghan Girl, Eiffel Tower and Mona Lisa are selected to be experimented with, and the compression errors are recorded together with image intensity, hard-threshold and storage saved. The results show that Afghan Girl and Mona Lisa have similar singular values behavior, but not Eiffel Tower. This study suggests that the rank must be at least 30% of the total image size to have good image quality and optimal storage use.