Improved one-dimensional dilation-based top-hat algorithm for star segmentation under complicated background conditions

Abstract

The white top-hat transformation has been widely used in small bright target extraction. It usually applies an erosion operation to remove the target and then a dilation operation to recover the intensity of the processed image. A bright target will be extracted by subtracting the opening operation (erosion followed by dilation) from the raw image. The drawback of this method is that its denoising ability is poor because the estimated background threshold by an opening operation is smaller than the raw image. This study puts forward the viewpoint that by use of a proposed one-dimensional (1D) symmetrical line-shaped structuring element a bright target can also be removed by the dilation operation. Consequently, the white top-hat transformation can be implemented by subtracting only the dilation operation from the raw image. To the best knowledge of the authors, it is the first time to use this method to achieve the top-hat transformation. The simulation experiment shows that the proposed 1D top-hat algorithm has excellent performance in denoising ability and detection ability. Moreover, real night experiments demonstrate that our proposed algorithm can work reliably under both complicated background conditions and good weather conditions. It is noticeable that the performance of computational efficiency and resource consumption have been considerably improved because a 1D structuring element is employed and the erosion operation is not included.