Research On Particle Swarm Optimization Algorithm For Denoising And Dehazing

Particle Swarm Optimization(PSO)is a swarm intelligence optimization algo-rithm and belongs to the important branch of Evolutionary Computation(EC).PSO has been widely used in image processing because of its advantages such as simplicity,less parameters,low computational cost,and fast convergence.This paper focuses on PSO for image denoising and dehazing applications.Denoising is a key step in image preprocessing.Bilateral filter is widely used in the field of denoising because of its“preserving and denoising" characteristics.The denoising effect of bilateral filter is determined by the value of its filtering parameters.In practical applications,these important parameters are often obtained through a large number of experiments,or are manually set,which affects the accuracy of experimental results.In recent years,researchers have proposed some automatic parameters selec-tion methods for bilateral filter,but they are usually complex,not universal or requiring statistical analysis of noised images in advance.To solve this problem,the standard P-SO algorithm is applied to the parameters selection for bilateral filter,and the bilateral filter optimized by the standard PSO is proposed.This method can automatically select the values of the bilateral filtering parameters for any image to be denoised,and then uses the optimized bilateral filter to denoise it.The use of the standard PSO improves the denoising effect and denoising stability of bilateral filter.The search accuracy of the standard PSO is poor.For it is easy to fall into the local extremum,it is not suitable for solving complex optimization problems in real-ity.To solve this problem,researchers have made many improvements of PSO,with the neighborhood structure being a major improving direction.However,most PSO improvements to the neighborhood structure do not take into account the authenticity and sociality of the structure.Therefore,this paper introduces the Random Apollonian Network(RAN),which conforms to the characteristics of real social networks,into the standard PSO,and proposes a new optimization algorithm RAN-PSO.The intro-duction of RAN improves the search accuracy and convergence speed of the standard PSO.Then,this paper applies RAN-PSO to image denoising,and proposes a bilateral filtering denoising method based on RAN-PSO.This method has stronger denoising effect and stability than the bilateral filtering denoising method based on the standard PSO.Dehazing is another important issue in the field of image processing.The dark channel prior algorithm and its improved versions are the most widely used dehazing algorithms currently,which have best dehazing effects.Similar to bilateral filter,the dehazing effect of this algorithm is closely related to the values of its parameters.At present,researchers select the parameters of the algorithm by trial and error.In ad-dition,due to the particularity of the foggy image and the difficulty in obtaining the image without fog,applying classical image quality assessment algorithms directly to the evaluation of the defogged image can not achieve good results.Therefore,this paper designes a new non-reference image quality evaluation algorithm as the fitness function,and then proposes a dark channel prior dehazing algorithm based on RAN-PSO.The method can automatically select the parameters of the dark channel prior dehazing algorithm for any foggy map,and then uses the optimized dark channel prior algorithm to dehaze it.The introduction of RAN-PSO can improve the dehazing visual effect of the dark channel prior dehazing algorithm.