Research On Contrast Enhancement Method For Single Image

Affected by the low-light environment conditions during imaging and the performance of digital imaging equipment,the captured images often exhibit low contrast.The previous single-frame image enhancement algorithms mostly adjust the contrast of the input image by adjusting the tone curve or performing histogram equalization.However,due to the limited information of the single image,these methods often fail to recover the detailed information of the image while enhancing the contrast of the image.The research content of this article is for the single image.Through image detail analysis and learning,and then use histogram equalization and deep learning ideas for image enhancement.The research content in this thesis is as follows:1.An image enhancement method based on the Multi-scale Fractional-order Hessian matrix(MSFr H)is proposed,which is an image contrast enhancement algorithm based on the weighted histogram of the Hessian matrix.This method can improve image smoothness.Suppress the contrast of the texture area while suppressing the area contrast.In the proposed method,the multi-scale fractional-order Hessian-matrix is firstly utilized to detect and quantify the texture information of the input image,which explores the regions that should be contrasted or should be restrained.Then,the strong texture regions are suppressed by a designed suppress function.Finally,the information on unsuppressed regions and suppressed texture regions will be count by a histogram,which is termed as Hessian-matrix weighted Histogram(Hess Hist)in this thesis.According to Hess Hist,the corresponding cumulative distribution function will realize the contrast enhancement operation on the input image.For real-time application,the integral images are introduced for fast computation of the Hess Hist.Experimental results show that the proposed Hess Hist-based image enhancement algorithm preserves the more details of input image without distortion,and is competitive with state-of-the-art image enhancement algorithms in both subjective visual perception and objective evaluation metrics.2.An image enhancement algorithm based on a supervised generative adversarial network is proposed.Firstly,U-net with global features is used as the generator of network to improve its ability to generate higher-quality images.Global U-net can extract global information to generate better results.The Patch GAN structure used by the discriminator can enhance the discriminative ability of the discriminator and have less parameters.In order to guide the network to learn more useful information and make the training process more stable,the loss functions use the L1 loss,the content loss of the pre-trained VGG network and the least square loss.The subjective and objective evaluation results of experiments show that this method is very effective for image enhancement.