An Auto-focus Method for Microscopic Images Based on QSOM Neural Network

Digital microscopes often require repeated manual focusing to obtain a clear image. Nevertheless, manual focusing easily results in artificial errors, making it difficult to evaluate the image definition, and the focusing process is slow and tedious. This paper presents an automatic focusing method based on Quantum Self-organizing Maps (QSOM) neural network and a new focusing evaluation function. The focusing evaluation function consists of Energy of Gradient (EOG) function and Discrete Wavelet Transform (DWT) function to better evaluates the sharpness of microscopic images at different focusing positions. The obtained data are used as training samples, and QSOM neural network is trained by focusing samples. The trained neural network can accurately predict the position achieve the focus position. Experimental results are given to verify the effectiveness of the proposed method.