In the image edge distortion correction,the straight-line projection- derived edge fit is poor,resulting in large correction errors.An image edge di stortion correction algorithm based on improved deep learning self-encoding is proposed.The adaptive threshold wavelet denoising algorithm is adopted to implem ent adaptive transformation of various scale parameters to complete noise remova l.According to the results of image denoising,Fisher vector coding is used to optimize the deep learning results,and the edge distortion shape of the image is extracted.And based on the edge distortion shape extraction results,the corr ection objective optimization function is obtained,the edge fracture is analyzed,and the linear projection derivative edge fitting is realized;by determining the image edge error evaluation function,the image edge distortion correction method is judged to achieve the image edge distortion correction purpose.The experimental analysis is carried out on noisy barrel distortion and pincushion dis tortion images.The results show that the proposed algorithm can effectively cor rect the barrel and pincushion distortion of the image.After the edge distortion of the barrel and pincushion image is corrected,the image The number of edges is the same as 5in the original image,realizing high-precision and high-efficiency image edge distortion correction.