The invention discloses a
JPEG (Joint Photographic Experts Group)
image compression artifact
elimination method based on a controllable
pyramid wavelet network, which comprises the following steps of: firstly, extracting an image feature related to a compression level, secondly, guiding to recover a Y channel image by using the feature, and then guiding to recover a CbCr channel image by using the feature and the recovered Y channel image, and finally, converting the image into an RGB space to obtain a final
recovery result. According to the method provided by the invention, image coding parameter information does not need to be predicted, a good
recovery effect can be achieved for a plurality of images with different compression levels, each
recovery network only needs to
train a single
network model, on one hand, the model uses jump connection to avoid the problems of gradient disappearance and gradient explosion possibly occurring in the training process, and on the other hand,
model complexity is reduced by using a
recursion module sharing parameter strategy, and efficient operation of the
algorithm is ensured. Therefore, the method provided by the invention has the advantages of simple model, few parameters, wide application range, remarkable
recovery effect and the like.