Pre-Processing of Panchromatic Images to Improve Object Detection in Pansharpened Images

In recent years, many techniques of fusion of multi-sensors satellite images have been developed. This article focuses on examining and improvement the usability of pansharpened images for object detection, especially when fusing data with a high GSD ratio. A methodology to improve an interpretative ability of pansharpening results is based on pre-processing of the panchromatic image using Logarithmic-Laplace filtration. The proposed approach was used to examine several different pansharpening methods and data sets with different spatial resolution ratios, i.e., from 1:4 to 1:60. The obtained results showed that the proposed approach significantly improves an object detection of fused images, especially for imagery data with a high-resolution ratio. The interpretative ability was assessed using qualitative method (based on image segmentation) and quantitative method (using an indicator based on the Speeded Up Robust Features (SURF) detector). In the case of combining data acquired with the same sensor the interpretative potential had improved by a dozen or so per cent. However, for data with a high resolution ratio, the improvement was several dozen, or even several hundred per cents, in the case of images blurred after pansharpening by the classic method (with original panchromatic image). Image segmentation showed that it is possible to recognize narrow objects that were originally blurred and difficult to identify. In addition, for panchromatic images acquired by WorldView-2, the proposed approach improved not only object detection but also the spectral quality of the fused image.