Context : Very few debris discs have been imaged in scattered light at wavelengths beyond \unit { 3 } { \micro \meter } because the thermal emission from both the sky and the telescope is generally too strong with respect to the faint emission of a debris disc .
We present here the first analysis of a high angular resolution image of the disc of \beta Pictoris at \unit { 3.8 } { \micro \meter } .

Aims : Our primary objective is to probe the innermost parts of the \beta Pictoris debris disc and describe its morphology .
We performed extensive forward modelling to correct for the biases induced by angular differential imaging on extended objects and derive the physical parameters of the disc .

Methods : This work relies on a new analysis of seven archival datasets of \beta Pictoris observed with the NaCo instrument at the Very Large Telescope in the L ^ { \prime } band , including observations made with the Annular Groove Phase Mask vortex coronagraph in 2013 .
The data analysis consists of Angular Differential Imaging associated with disc forward modelling to correct for the biases induced by that technique .
The disc model is subtracted from the data and the reduction performed again in order to minimize the residuals in the final image .

Results : The disc is detected above a 5 \sigma level between 0.4 \arcsecond and 3.8 \arcsecond .
The two extensions have a similar brightness within error bars .
We confirm an asymmetry previously observed at larger distances from the star and at shorter wavelengths : the isophotes are more widely spaced on the north-west side ( above the disc apparent midplane ) than on the south-east side .
This is interpreted as a small inclination of the disc combined with anisotropic scattering .
Our best-fit model has an inclination of 86 ^ { \circ } with an anisotropic Henyey-Greenstein coefficient of 0.36 .
This interpretation is supported by a new asymmetry detected in the disc : the disc is significantly bowed towards the north-west within 3 \arcsec ( above the apparent midplane ) .
We also detect a possible new asymmetry within 1 \arcsec , but at this stage we can not discern between a real feature and an underlying speckle .

Conclusions :