Shading correction methods for digital image analysis of confocal wood images.

摘要

Confocal fluorescence microscopy provides a rapid method for acquiring high quality optically thin section images of wood suitable for measurement of cell dimensions. Single optical slice images of wood may occasionally contain artefacts due to differential light absorption caused by variation in the distance between the sample surface and the imaging plane across the field of view. Regional brightness variations, which we call shading, may cause problems when such images are used for wood cell measurements using digital image analysis, affecting the accuracy of wood cell dimensions. We have compared various shading correction methods for confocal microscope images of Pinus radiata wood and investigated the effect of shading on both the classification of cell wall pixels and the resulting cell dimension measurements. Severe shading results in significant errors for measurement of cell wall area, but smaller errors for cell wall thickness and lumen diameter. Some shading correction methods have unwanted effects on pixel classification and cell dimensions, while more effective methods remove the shading without introducing further artefacts. The effect of shading is influenced by choice of thresholding method.