Adaptive low-gray image enhancement based on BP neural network and improved unsharp mask method

摘要

In this paper, an image enhancement algorithm based on BP neural network and improved unsharp mask method is proposed to solve the problem of the unclear image detail information and the fuzzy edge information after wavelet decomposition. Firstly, the variance, mean value and information entropy of the original image are obtained. At the same time, wavelet decomposition is applied to obtain the low-frequency and high-frequency images of the two layers. Then the improved image model based on unsharp mask method is established by virtue of high and low frequency images. BP neural network is employed to predict and reconstruct the processing coefficients of the obtained image model according to the variance, mean and information entropy of the image. Finally, the enhanced image is obtained by histogram equalization. Experiments show that this method can improve the contrast of low-gray image effectively and solve the problem of image blurring, and make the final image have good visual effect.