Compact color descriptor for object recognition across illumi- nation changes

摘要

In this paper, we propose a compact color invariant image descriptor which characterizes both the color distribution and the spatial interactions between the pixels in the image. Our approach is based on the analysis of the pixel rank measures which are their ranks when they are sorted according to their color component levels within a color image. Indeed, we show in this paper that the correlation between the rank measures of neighbor pixels in a color image is an efficient feature to describe the content of this image. This descriptor, extracted from the chromatic co-occurrences matrices, has the advantages to be invariant to illumination changes, low-time consuming, highly discriminating and compact. The proposed rank correlation coefficient is used by our object recognition scheme whose effectiveness is assessed with a public database that contains images of objects acquired under different illuminations.