Patch-Based Image Learned Codec using Overlapping

End-to-end learned image and video codecs, based on auto-encoder architecture, adapt naturally to resolution, thanks their convolutional aspect. However, while coding high resolution images, these codecs face hardware problems such as memory saturation. This paper proposes a patch-based solution an end-to-end model, which aims remedy the limitation maintaining same quality full coding. Our method consists in overlapping patches of reconstructing them into decoded using weighting function. approach manages be par with performance endto-end even slightly outperforms it, being adaptable different sizes. Moreover, this work undertakes study effect patch size solution’s performance, consequently determines best terms time efficiency. Finally, introduced is also compatible any codec conv/deconvolutional autoencoder architecture without having retrain model.