Statistical models for quality criteria of image compression algorithms (Q2722259)

Image compression deals with the problem of efficient image coding in such a way that redundant bytes may be suppressed without loosing relevant information for recovering the original human visual quality. Hence, considering an image as a data table and using statistical techniques for reducing dimensionality may be helpful. The main purpose of this paper is to develop statistical models that allow to define new efficient quality criteria for image compression algorithms.NEWLINENEWLINENEWLINETo develop these new criteria several statistical approaches are used. For testing purposes, four image compression or transformation algorithms are used: discrete cosine (DC), fast Fourier (FFT), wavelets (WAV) and fractals (FRAC). Each of these algorithms has different time- and compressing-efficiency, but this paper is only interested in developing new criteria for measuring the image quality of the recovered image after compression. Starting with a unique image, the classical Lena, three to four compression levels are used for each of these algorithms. Moreover, in addition to the four compression algorithms, the most commercially popular called, JPEG, is used which is actually also based on the DCT.