Stereo matching is the process of finding corresponding points in two or more images. The description of interest points
is a critical aspect of point correspondence which is vital in stereo matching. SIFT descriptor has been proven to be better
on the distinctiveness and robustness than other local descriptors. However, SIFT descriptor does not involve color
information of feature point which provides powerfully distinguishable feature in matching tasks. Furthermore, in a real
scene, image color are affected by various geometric and radiometric factors,such as gamma correction and exposure.
These situations are very common in stereo images. For this reason, the color recorded by a camera is not a reliable cue,
and the color consistency assumption is no longer valid between stereo images in real scenes. Hence the performance of
other SIFT-based stereo matching algorithms can be severely degraded under the radiometric variations. In this paper,
we present a new improved SIFT stereo matching algorithms that is invariant to various radiometric variations between
left and right images. Unlike other improved SIFT stereo matching algorithms, we explicitly employ the color formation
model with the parameters of lighting geometry, illuminant color and camera gamma in SIFT descriptor. Firstly, we
transform the input color images to log-chromaticity color space, thus a linear relationship can be established. Then, we
use a log-polar histogram to build three color invariance components for SIFT descriptor. So that our improved SIFT
descriptor is invariant to lighting geometry, illuminant color and camera gamma changes between left and right images.
Then we can match feature points between two images and use SIFT descriptor Euclidean distance as a geometric
measure in our data sets to make it further accurate and robust. Experimental results show that our method is superior to
other SIFT-based algorithms including conventional stereo matching algorithms under various radiometric changes.