2. Histogram Matching

Last chapter we discovered that we cannot make sure the shape of the science fiction style image's histogram is U-shape, but we can see the similarities of the science fiction style image. In this chapter, we would like to use the technique calls histogram to take advantage of the image histogram we saw from last chapter.

(For the below images, we would use gray-scale image for generating the histogram and the output, we will talk about coloring the gray-scale image in the next chapter)

Histogram matching is an algorithm to force one image's histogram to be similar (as close as) the reference one. For example, here is the following reference image and the source image.

image1.1) reference image

image1.2) source image

originally, these are the histogram for both images:

image1.3) histogram for reference image

image1.4) histogram for source image

Our goal for the histogram matching, is to force the histogram in 1.4 to be as similar as possible compare to 1.3

It is not possible to be exactly the same, since the pixel distribution is difference if you force it to be exactly the same it might result in a weird image. After that, we can get this histogram

image1.5) new histogram after histogram matching

For the result, we got this image according to image1.5 histogram

image1.6) output after image histogram matching

This is a good result that we can see since it makes the darker place becomes darker and light becomes brighter, which is a good tiny step on making science-fiction style image.

However, as we already discovered, different kinds of science-fiction style image, will have different kinds of histogram. Therefore, having more reference images for generating a science-fiction style image will be important for our result.

Here, we arranged the order of multiple references by the percentages of the light source. It might not be accurately arranged but we might use some algorithm to detect how much light sources in the source image to select which reference image we should use.

image2.1

image2.2

image2.3

image2.4

image2.5

image2.6

image2.7

image2.8

image2.9

image2.10

We can try to use the light detection to select which reference image we use, or we can let the user select which style they think is the best kind of style they want for their science-fiction style image.

Image credit:

https://visualcocaine.org/photo/1826/fantasy-planet-archer-woman-moon

http://www.diazmag.com/explorers-court-metrage/

https://www.mapmaker-malique.de/blog/gebaerende-kinder-und-gesellschaft/

https://www.wallpaperup.com/1081091/naves_futuristas_sci-fi_aliens.html

http://blogs.adobe.com/creative/visual-trends-rueckblick-2017/

https://pixabay.com/images/search/science%20fiction/

https://www.this-is-cool.co.uk/the-sci-fi-concept-art-of-adam-kuczek/

https://www.freelancer.com/contest/Scifi-Human-vague-concept-art-1055269-byentry-15178777

http://hdqwalls.com/scifi-concept-art-fantasy-wallpaper

https://www.msn.com/nl-be/entertainment/film/netflix-koopt-rechten-op-chinese-megahit-the-wandering-earth/ar-BBTVEFP