Liquid phase transmission electron microscopy allows the imaging of materials in liquid environments. The sample is encapsulated within electron-beam transparent windows and hence protected by the ultrahigh vacuum necessary within the electron gun. Such an approach allows to study biological and soft materials in their natural environment and offers the possibility of accessing their dynamic nature. Yet, the electron beam scattering from the windows and solvent increases the image noise and blur. Herein, we propose a pipeline to both de-noise and sharpen images obtained by liquid transmission electron microscopy. We develop the workflow in a way that it does not require any human interference, nor introduce artefacts, but actually unveils features of the imaged samples covered by the noise and the blur. LAY DESCRIPTION: Transmission Electron Microscopy TEM is one of the most powerful techniques for structural determination at the nanoscale, with the ability to image matter down to the atomic level. TEM is only possible by keeping the electron beam under high vacuum in order to avoid undesired scattering events in the beam path. High vacuum means that the TEM samples must conventionally be in solid-state. Thus, samples in liquid form or containing liquids, like water, need special preparation techniques which tend to alter the structure and chemical nature of the sample. Such alterations are particularly critical for biological and soft organic materials where the structures are controlled by the presence of water and/or other liquids. The development of new cameras, materials and sample holders have made possible for TEM to be performed on liquid samples. Liquid Phase Transmission Electron Microscopy (LTEM) offers the possibility to investigate nanoscopic structures in liquid state and monitor dynamic processes. However important limitations come from the liquid nature of samples in the imaging process such as the low contrast afforded by organic and biological materials and additional noise and blur introduced by the liquid sample and its thickness. Existing image analysis algorithms for TEM result inadequate for LTEM. The end-to-end image analysis method herein has the ability to recover the original images together with their sharpness, without introducing any artefacts. The proposed algorithms offer the great advantage of unveiling image details which are not usually seen during imaging, thus allowing a better understanding of the nature, structure and ultimately the function of the investigated structures. The fully automatised analysis method allows to efficiently process dozens of images in few hours, improving dramatically the performance of LTEM imaging.