We point out the interest in using Zipf law in the wide field of image analysis. We show how it is possible to adapt it to image analysis. There are many applications; here we are concerned with image compression and focus on a new definition of quality measurement of an image obtained after a decompression process. First we propose a coding mode of an image using some neighborhood of each pixel and the Zipf law. Then some significant comparisons can be achieved between different images. A statistical study of the patterns present in the image allows a characterization of the information contained in the image so that a measurement of information modification is made possible. We define two quantitative parameters; one indicates the structure of the image whereas the second indicates high content information. The examples of the images of a landscape and of a female face are used to illustrate the influence of the compression rate.
展开▼
