With histogram matching we can transform an image's intensity distribution. But this idea is very subjective. Sometimes for choosing a target image we have options no better than a guess. But is their any rule of thumb of choosing an appropriate target image for desirable improvement in input image? Is it possible to construct an artificial image and use it as a target image to obtain a specified improvement in input image? Is there any mathematical approach to obtain this? How things will change for a color image?

  • $\begingroup$ Unless you are trying to account for predictable differences between two images of the same scene, you don't need a second image. What is the particular application you are working on? $\endgroup$ – A_A Jun 26 '20 at 8:05
  • $\begingroup$ @A_A trying to make low light images (having poor contrast) more visible. $\endgroup$ – hafiz031 Jun 26 '20 at 11:09
  • $\begingroup$ ...then why not just add a simple window so that you can map the regions that are now in low grey values to lighter regions? $\endgroup$ – A_A Jun 26 '20 at 11:30
  • $\begingroup$ @A_A increasing low light image's visibility is just a part of my whole work. Again it is not the case that all of the images will be low-light images. More specifically, I am making a real-time object detector where in some cases low light captured frames may come, so I have no way to process them manually or offline. The whole process must be in real-time. So this is the real problem. Another problem is histogram matching is not fast enough to be able to handle real-time frames. $\endgroup$ – hafiz031 Jun 26 '20 at 11:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.