Research On Enhancement Algorithms Of Low Illumination Images Based On Non Subsampled Shearlet Transform

As we all know,due to the influence of factors such as poor lighting environment or equipment limitations,various problems occur in the process of image formation or transmission,such as reduced contrast,loss of details,and severe color distortion.Under the circumstance,effective information can't be obtained from those images,and it is impossible to subsequently perform the image analysis and recognition.Therefore,the technology of low-illumination image enhancement has great research value and prospect.This thesis studies various algorithms for image enhancement and analyzes their advantages and disadvantages,and then proposes improved algorithms based on the characteristics of the images captured in a low-light environment.The proposed algorithms strive to overcome the defects of other algorithms and improve the image quality to better fit the human visual system.The specific research contents of this thesis are as follows:1.This thesis proposes an enhancement algorithm for high signal-to-noise ratio images based on non-subsampled shearlet transform.The HSV color space is more in line with the human visual characteristics and has lower coupling,so the proposed algorithm in this thesis is performed in the HSV space.Histogram equalization and non-subsampled shearlet transform are first conducted for the V channel.Next,gamma correction is performed on the obtained low frequency components,and improved guided filtering is performed on the high frequency components.Finally,the V channel is reconstructed by combining with the H and S channels,and converted back to the RGB color space to obtain the final enhanced image.Experimental results show that the proposed algorithm can not only effectively improve the contrast and brightness of low-illumination images,but also highlight more details and higher color fidelity.2.This thesis proposes an enhancement algorithm for low signal-to-noise ratio images based on non-subsampled shearlet transform.In the Lab color space,Gaussian white noise has little effect on a and b channels,so the proposed algorithm in this thesis will be performed in the Lab space which is more conducive to remove noise.An image is first converted from the RGB color space to the Lab color space,then the color components a and b are processed by bilateral filtering,and the luminance component is subjected to non-subsampled shearlet transform.Gamma correction is performed on the acquired low frequency components to enhance the contrast,and soft threshold denoising is used on the high frequency components.Finally,it is converted back to the RGB color space and performed white balance to correct the colors to obtain the final image.Experiments show that in the subjective aspect,the images processed by the algorithm in this thesis show more details and better visual effects;in the objective aspect,the proposed algorithm can remove noise more effectively,and the enhanced images retain more image features and have higher feature similarity with the original images.