PSF estimation for Gaussian image blur using back-propagation quantum neural network

During spatial remote sensing imaging procedure, combined degradation factors conduce to Gaussian image blurring. The precondition of restoring the degraded image is to estimate point spread function (PSF) of the imaging system as precisely as possible. Because the depredating processes are quite complex, the transfer function of the degraded system is often completely or partly unknown, which makes it quite difficult to identify the precise PSF. Considering the similarity between the quantum process and imaging process in the probability and statistics fields, a novel algorithm is proposed by using multilayer feed-forward back-propagation quantum neural network (QBPNN) to estimate PSF of the Gaussian degraded imaging system. Different from the classical artificial neural network (ANN), 2 adjustable parameters of weight connection coefficient and phase coefficient are introduced in its quantum neurons used in learning stage. By establishing different training sets, this estimation method can overcome the limitation in the dependence on initial values and large amount of computation. Test results show that this method can achieve higher precision, faster convergence and stronger generalization ability comparing with the traditional PSF estimation results.