Entropy optimized contrast stretch to enhance remote sensing imagery

This paper presents a contrast stretch (CS) method based on minimum entropy constraint to enhance images obtained in remote sensing applications such as ground penetrating radar (GPR), synthetic aperture radar, and infra-red imagery. The CS enhances contrast of the low-contrast part of an image. In remote sensing, it is usually the desirable signals that are of low contrast while interference of high contrast. The CS modifies the original image such that pixel values above and below preset boundaries are set to Zero and the maximum possible pixel value and the pixel values falling between the boundaries are stretched out to enhance the contrast of the image. Using the CS we can enhance the contrast of desirable signals from, for example, a buried landmine or an object obscured by some interference. On the other hand, the CS inevitably enhances other parts of a remote sensing images, such as clutter and measurement noise. Therefore there is a trade-off in using the CS. It is beneficial to find the correct "cut-off" boundaries in the CS in some optimal sense. We propose using minimum entropy as a criterion of looking for the optimal CS parameter Using field data from GPR application, we show that improved image can be obtained which makes further processing such as detection more accurate.