Abstract:
This paper proposes a novel method for human
detection from static images based on pixel structure
of input images. In training stage, all sample images
consisting of human images and non-human images
are used to construct a Hausdorff distance map based
on statistically analyzing the difference between the
different blocks on each original image. A projection
matrix will be created with Linear Discriminant
Method (LDM) based on the Hausdorff distance map.
This projection matrix will be used to transform multidimensional
feature vectors (distance maps of testing
images) into a feature in a one-dimensional domain.
The decision will be made on the simple onedimensional
feature domain according to a precalculated
threshold to distinguish human figures from
non-human figures. In comparison with the common
method based on Mahalanobis distance maps, the
proposed method based on Hausdorff distance maps
performs much better. Encouraging experimental
results have been obtained using 800 human images
and 800 non-human images.
