Image quality assessment in time/frequency surfaces

In this paper time-frequency analysis has been done using the traditional short time Fourier transform (STFT), also known as a spectrogram, and the short time fractional Fourier transform (STFrFT). The STFrFT is a generalisation of the traditional STFT. Performance comparisons are made between the two different transforms and it is shown that matching the transform to the nature of the data improves the quality of the time-frequency images. Two data sets are analysed, both sets were recorded in the Baltic Sea. The first set includes high bandwidth (60 kHz) high chirp rate, short duration signals (1 ms duration). The second set has lower bandwidth and lower chirp rate. The analysis of these two sets are used to illustrate the strengths and weaknesses of the transforms. In the second section of this paper a number of different metrics for image quality assessment are tried out on the time/frequency images produced with the aforementioned transform techniques. The results show that the FrFT has several advantages over the traditional short time Fourier transform, as it makes use of the knowledge of the transmitted wave form, when the data indeed has the desired properties. When the chirp rate decreases, the FrFT becomes the traditional FT and does not provide any enhancement to the time/frequency image. The implementation of image quality measures show that depending on what the characteristics of the features in the image are, different image quality measures are required to ensure that the right conclusions are drawn about image quality. This also points out the need to establish what one means with image quality.