Last modified: 11-05-2021
Abstract
Image comparison is widely used in application software. Modern methods of comparison allow you to compare and analyze images and draw conclusions from the data. There are methods that generate so-called hash codes, ie 64-bit numbers that characterize the image. Other methods allow you to detect, read and analyze individual features of the image. This paper proposes the use of a discrete cosine transform, the result of which is used to calculate the hash code of the image. This method is effective for comparing slightly modified images (up to 25% of the modification area), is quite simple to implement and effective in speed and insensitive to any image modifications. Simple hash comparison methods have the same advantages, but are very sensitive to any image modifications The algorithm for comparing images using a hash code is reduced to obtaining the value of the hash code, which can be compared with other hash codes using the Hamming distance. When determining the limit value of the Hamming distance, it is possible to state the similarity or difference of the images.
The algorithm can map images even if one of them is resized and if the aspect ratio of the image is changed. Small changes in color are also allowed (brightness, contrast). Cropping and splitting images can increase the efficiency of calculations both in terms of speed and accuracy of the result.