A New Local Enhancement Algorithm for Small Target Detection Based on Top-hat Transform

One of the main breakthroughs in the area of machine visualization is the tiny objective exploration. Several techniques have been put out in recent years for locating small infrared targets. Aiming at the problem of missing detection in infrared faint and marginal objective exploration, and to improve the effect of detection, a top-hat transform based on a local enhancement algorithm is proposed. In the first stage, the characteristics of the top-hat transform are analyzed, and then the target enhancement algorithm is introduced. The proposed local enhancement algorithm uses the square of the ratio of the pixel mean of the objective region to average grayscale of the backdrop area as a weighted term. To better highlight objective while suppressing backdrop, the background area is divided into multiple identical blocks, and then the square of the proportion of mean value of pixels in objective region to the largest mean of the backdrop block is used as the weighting term, which greatly enhances the significance of the target. In the second stage, top-hat transform was adopted to detect images that have been processed. Finally, significance image was segmented by threshold value to obtain the real target to be detected. When the target is very weak, this method can enhance the weak target, which is beneficial to detect the target.