Research On Adaptive Enhancement And Restoration For Degraded Images

With the popularization of digital imaging devices,massive digital images or videos are emerging in various applications.Due to the influence of the illumination conditions,the atmospheric environment and the imaging devices,the quality of the acquired image often does not meet people’s expectations and technical requirements.Therefore,it is of great practical significance to study adaptive enhancement and restoration of degraded images.Based on the imaging model,this paper focuses on three types of degraded images with uneven illumination,fog and low illumination.The existing algorithms have many limitation,such as the parameter setting according to the image alone,one algorithm only dealing with a single degraded type and many quality indicators difficult to balance.This paper improves and optimizes the algorithm from the aspects of adaptive processing,extension algorithm to deal with multiple degradation types and consideration performance index,which improves the processing performance of algorithms.The main research work of this paper are as follows:(1)In terms of adaptive processing of uneven illumination images,spatial multi-scale adaptive homomorphic filtering and adaptive local entropy guided image filtering are proposed aiming at the over-enhancement effect of homomorphic filtering.In this paper,we propose a spatial multi-scale algorithm to improve the homomorphic filter to avoid over-enhancement,and then design an adaptive algorithm for the cut-off frequency to improve the stability of the algorithm.Finally,an adaptive local entropy guided image filtering removes blocking artifacts,which can take both remaining edge and reducing noise into account.This method is processed on the V-channel in the HSV color space to obtain a uniform illumination and vivid color enhancement result.(2)In terms of adaptive sharpening of the foggy image,an adaptive image scene-based restoration algorithm is proposed,which is on the basis of haze removal using the dark channel prior.First of all,aiming at the inapplicability of the prior in the sky region,this paper proposes an adaptive sky region detection segmentation algorithm,and divides the image into distant sub-images and close sub-images.Then according to characteristics of different scene sub-images,the adaptive transmission estimation and optimization algorithm is designed,and the morphological opening operation is used to improve the dark channel prior.Finally,an adaptive interval estimation method is proposed to determine the atmospheric light value.This method can obtain brightly colored and clearly restored results under different image scene conditions.(3)In terms of adaptive processing of low-illumination images,this paper studies the extended application of dark channel prior haze removal in low illumination images,and proposes a fast restoration method combined with noise reduction.Firstly,based on characteristics of low-illumination inversion images,the V channel is proposed to use adaptive estimation of transmission,which saves time overhead without optimization.Then the atmospheric light can be estimated adaptively to get a fast recovery method.Finally,considering the noise amplification problem in degradation model,an adaptive improved dual domain filtering algorithm is proposed to deal with it.This method can quickly obtain the restored results of high contrast and moderate brightness.