Adaptive Light Field Contrast Enhancement using Efficient Feature Extraction

Light field (LF) image can provide comprehensive image information and real visual experience. In this paper, we propose an adaptive enhancement method of LF sub-aperture images (SAIs) based on machine learning. The contrast of SAIs is enhanced with scene adaptive mapping curve. Firstly, the data set representing image features is extracted from dominance SAI. Then the multi-layer perceptron (MLP) is used to train the data set to obtain adaptive mapping curve parameters, thereby adjusting the image ulteriorly and enhance the contrast. Experimental results show that the predicted adaptive parameters using our network has high accuracy and is close to ground truth. Depending on comparing with other existing image contrast enhancement, the results demonstrate that the global image contrast is improved without over-enhancing the local region. The proposed method reduces the complexity of image processing and realizes real-time adaptive optical field enhancement.