Frequency Domain Image Representation in Blocks Form for Textures Comparision
A. Shoberg, Sergey Sai, K. Shoberg · 2019 International Multi-Conference on Engineering, Computer and Information Sciences (SIBIRCON) · 2019
The proposed approach is used to store images in compressed form. At the same time, elements of invariance to rotations are preserved. It's described block frequency transforms executions on example of the discrete cosine transform which using for images compression. Two execution variants lead to invariance to process direction. Transform results are used to compare images including textures. The comparison results for textures base are presented. Better results are explained by the expansion of correlation coefficients dynamic range compared to the traditional approach to image representation. The proposed approach shows that the computational complexity in the transform implementation does not change, both for the standard algorithms and for their fast implementations. This is due to the fact that for each block the calculations are made independently. The number of multiplications and additions remains the same. It changing only the order of elements reading and recording for calculations. The proposed approach can be used for other integral transforms.