A.Yu. Zrazhevsky, A.V. Kokoshkin, E.P. Novichihin, S.V. Titov

Methods for the improvement of the spatial resolution [1] are very important in the processing and restoration of image quality. In particular, the pictures obtained by antennas operating in the millimeter wave band have restricted resolution due to small antenna dimensions. Here, a method for the improvement of thermal radio image quality in the millimeter wave band is discussed. The theory of upward filtration is applied in the processing of the Fourier spectrum of the image. It is known [1] that the ef-fect of antennae in receiving a radio image is to produce a modulation (filtration) of the spatial spectrum of the image. Thus, for the improvement of the spatial resolution, we consider an approach based on a knowledge of the hardware function system (HFS) used in receiving the radio image. By knowing the HFS and the Fourier spectrum of an etalon image one can restore the spatial spectrum of the original image and hence the picture of the detected object. Numerical simulation is used to demonstrate the effectiveness of the proposed approach. The spectrum of etalon image allows eliminating of spurious amplitudes in the spectrum of restoring image. These spurious amplitudes arise in the recovering procedure due to division by small numbers. Amplitudes in the small regions of the etalon spectrum are used in order to establish a threshold criterion for amplitudes of the recovering spectrum and to determine filter parameters. The subsequent filtration allows one to eliminate moiré in the resultant image and significantly enhance the spatial resolution. Numerical simulation is firstly used to deteriorate optical image in full analogy with the real radio image [2]. Different values of the restoring filter (Gauss filter) width ∆ are used to estimate the influence of inaccuracies in the determination of the HFS. The difference between the restored image and the original image is small if ∆=0.95-1,05∆0 Here ∆0 is the degradation filter width corresponding to original form. The restoration of the image is almost impossible if ∆>1.2∆0 or ∆<0.8∆0. So, the acceptable error in the restoring filter width is approximately not higher than 10%. The improvement of image quality is still possible in this case. Concerning the investigation of the noise effect, a model noise is added to the image in such a way that noise influences each pixel independently. The noise has normal (Gauss) distribution and varied standard deviation from the mean. The noise effect is insignificant if «signal to noise» ratio > 30 dB. The quality of the restoring image notably decreases if «signal to noise» ratio < 25 dB. The influence of the image scale was also investigated. The spatial resolution is not changed significantly if the scale difference between restoring image dimensions and etalon image dimensions is in the range 0.3-1.5. However the contrast range of the image is changed in this case. Numerical simulation shows that filtration of spurious amplitudes of Fourier spectrum of the restoring image based on usage of the etalon spectrum and on a knowledge of the HFS receiving the radio image (with 10% accuracy) permit improve the spatial resolution of image if signal-to-noise ratio is more than 25 dB. The proposed method of image restoration allows one to improve the image quality.