WTEC: a wavelet transform–based image exposure correction method

Captured images may suffer from underexposure or overexposure, due to various factors, including environmental lighting conditions and imaging devices. This can lead to poor image quality and affect the accuracy of downstream machine vision tasks. Previous research has focused primarily on correcting underexposed images, which does not address the actual needs of real scenarios. In addition, most of these methods operate in the spatial domain, which makes it difficult to correct underexposed or overexposed images simultaneously. Therefore, a wavelet transform–based image exposure correction method is proposed to improve the image correction quality. This method processes images in the frequency domain instead of the spatial domain and corrects underexposed and overexposed images simultaneously. Specifically, the underexposed or overexposed image is first decomposed into one low-frequency sub-image and three high-frequency sub-images through wavelet transform. Then, the low-frequency sub-image is corrected for the brightness using a well-designed U-Net network, whereas the high-frequency sub-images are enhanced for the texture details using residual networks. Finally, the corrected low-frequency and high-frequency sub-images, processed by the network, are reconstructed into a corrected image using an inverse wavelet transform. The effectiveness of the proposed method is validated by both quantitative and qualitative experimental results.