Adaptive enhancement of image brightness and contrast based on neural networks

A novel adaptive image brightness and contrast enhancement method combining high-frequency emphasis with neural networks is proposed. First, the low frequency components of the image are obtained by average filter, then the high frequency components of the image can be obtained by subtracting the low frequency components from the original image, and the nonlinear mapping relation between the enhanced factors of image brightness and contrast, the mean and standard deviation of the original image is established based on neural network. The weighting factors are automatically determined by the constructed neural network in terms of the mean and standard deviation of the image. The new algorithm has very small computational complexity while still produces high contrast output images especially for low-intensity and low-contrast images, which makes it ideal to be implemented for on-line detection system based on dynamic image process. The proposed technique is tested in the images collected by trouble of moving freight car detection system(TFDS), and a very good result has been obtained.