Then, the images were recorded with an acquisition
data software (Visilog 5.4) and represented in 256 grey
levels. Afterwards, they were converted from pixels to
micrometers, depending on the magnification selected.
Due to problems to separate correctly the three main
phases during the image thresholding and to create a
representative skeleton, the image analysis was only
carried out by tracing with a mouse. Due to irregular
shapes ofthe ice crystals, we measured two characteristic
lengths ofthe ice crystal to evaluate the particle
area, in order to calculate the equivalent circle diameter.
For each sample, about 200 crystals were analysed and
the size distribution curves were plotted with an
increment diameter Dd ¼ 10 mm. Air bubbles, as expected,
appeared as dark zones and were ofquite
spherical shape. Ice crystals appeared in grey colour and
the cryoconcentrated continuous phase (containing the
dry matter and the unfrozen water), in white levels. Fat
globules—with mean diameter values according to
literature around 1 mm—could be the dark points round
about the air bubbles surface (Fig. 2).were not taken into account satisfactorily, particularly,
due to the low maximum magnification ofour stereomicroscope.
For the lower magnification, the ice crystals
average diameter was higher than the corresponding
values obtained with higher magnification. Finally, to
avoid artefacts due to image analysis limitations, we
adopted and kept for all the experiments analyses, the
same image magnification, namely X ¼ 100; except in
the case oftoo large ice crystals sizes.