Fantastic work!
I’ve been playing around with V1.1 over the weekend and found some shots that are hard to match. Either they don’t match the colors perfectly or if they do there is too much noise. So it’s great that you tackled this problem with the new version. I also found that with some shots you only need just a few rotations
I have implemented a simple algorithm to tackle the dark colors of the blu ray. It goes like this: Calculate the average red green and blue value of both the input and the target. Subtract the average input values from the target values. Add the differences to the red green and blue values of each pixel in the input. I used the result image of my previous post as the input and the screener image from your initial post as the target.
The first result was this: https://s31.postimg.org/81lhm6g89/adjustedoriginal.png
If you compare this with the original screener you can see that the result is to bright. If i just add half the difference I get this result: https://s32.postimg.org/7s3ffif8l/adjusted.png
This is already really close to the screener. But I think it makes sense to introduce a parameter value in order to control the difference to add.
I remember trying to transfer the colors from the Screener to the Blu Ray a year ago. I used a similar approach as DrDre does for his program. However, I ran in the same issues that you described in your initial post. The main problems are the crushed whites and blacks of the source. I cancled the project out of frustration.
Your approach of combining the lumina of the Blu Ray with the chroma of the Screener got me motivated again. The colors of the screenshot are really good. The only issue is the color smear. You can see it around the suit of the Joker.
I wrote a simple program to fix this issue. The parameters of the program are BluRay_EE.jpg (input) and Blu_Ray_w_Screener_Chroma.jpg (target). I compared the rgb value of each pixel. For each rgb value of the input I got a list of rgb values of the target. For most input rgb values, the size of the list is just one. However, even if the size is greate than one, the rgb values are pretty similar. I calculated the average red green and blue value based on the list of target values. This resulted in a map that maps each input rgb value to exact one rgb value. Based on this map I recolored the input image, which resulted in this image:
Link https://s32.postimg.org/4xeg76khv/mapped.png
As you can see the color smear is almost gone. There is also a slight color change, which in my subjective opinion doesn’t really hurt.
I am really pleased with this approach. The next step is to find a way to estimate color pairs that aren’t in the color map. Once this is accomplished, you are able to calculate a color map based on a few frames and can use it for other similar frames. By doing this you would be able to change the colors of the IMAX scenes.
