Automated visual inspection tasks are frequently concerned with the ex
amination of homogeneously textured surfaces such as fabrics, wallpape
rs, machined surfaces, and floorcoverings. Often, the images taken fro
m such surfaces are degraded by an intensity inhomogeneity due to the
image acquisition process. This inhomogeneity is considered to be an i
rrelevant and disturbing signal component, which should be suppressed
to enhance the desired texture component and to ease a subsequent text
ure analysis. We show that, especially for textured surfaces, it is no
t always reasonable to assume a pure multiplicative composition of the
texture signal and a disturbing inhomogeneity. We introduce a notion
of homogeneity of n'th degree based on first-order statistics and pres
ent image processing methods for the homogenization of first, second,
and infinite degree. For the homogenization of second degree, we propo
se a computationally efficient frequency domain signal processing meth
od with high homogenization performance and low nonlinear distortion.
Furthermore, we suggest a high-performance homogenization of the infin
ite-degree technique that equates the local histograms to a global his
togram, which is adapted to the given image data. We compare the propo
sed homogenization methods visually and quantitatively with the well-k
nown homomorphic filtering technique, which assumes a pure multiplicat
ive inhomogeneity. We demonstrate that our methods achieve much better
results for synthetic as well as for realistic images of textured sur
faces. (C) 1997 Society of Photo-Optical Instrumentation Engineers.