As some of you may have seen, I have been working on measurements of beam profiles using a lux meter, and presentation methods for the output for some time. I have taken measurements out from the center of the hot spot and radially outwards in several directions to get a representative result.
Earlier this month, I read this tread: Beam Reviews - Update 2/18/07 (added D-mini - also added/changed plots) and got very interested in the suggestion from
656nm at the end of the tread.
Since that, I have been investigating on how to use my camera to get the full output profile from my lights. The idea was to get the lux value in some million points by just one shot.
I painted a wall white with the most diffusive paint I could find, and made some marks as reference points. I put up my L2D at 1 m distance from the wall, took measurements at all reference points, and started to take photos.
I soon found out that it was not that easy as I had hoped. I could only get out the light level in 255 different steps from the photos. This is of course far to little if you want to measure levels of several thousand lux and down to a few. This means i have to take several photos with different exposures.
Another difficulty is that the relationship between light level in the photos and the lux value seems to be linear only for quite low levels. If the light level comes above 100 in the photo (255 is max) then the relationship becomes quite non linear. Regions of the photos with too low light level is difficult to use because a step of one level may mean a big difference in lux level.
I searched around to try to find image processing software that could help me with these problems, and I tried several, but non of them could give me what I wanted, and became very cumbersome to use. So I decided to take up old skills and make the analysis software myself (I worked a lot with programming up to about ten years ago).
I have based my code on the package called ImageMagick and the interface towards the C programming language called ImageWand.
The program I have written, reads in several photos, and compares them with the lux readings from my lux meter. It generates some tables, and writes out a lot of data indicating the accuracy of the conversion of the photos into lux values. There have been a lot of thinking, analysing and trials and error before I could get this to work, but now it does. Here are an example of what I can get out:
This picure is generated out of 9 different images shot with exposure times from 1 s down to 1/1000 seconds. In this picture, every pixel is proportional with the lux value except that I have increased the pixel values at the lower levels just to make them visible. Another alternative is to use a logarithmic scale for the lux levels.
There are a lot of details in this picture that is difficult to see since I have had to reduce to an allowable size. If you see closely, you can se that the rings in the center is skewed slightly to the left. I suspect this is due to specular reflection from my white wall. I will try out other white surfaces than the painted I have now to see if I can get rid of it.
When I am satisfied with my generated lux pictures, then I can take out any profile, and can of cource also very easily calculate the lumens value. I will then figure out the best ways of presenting the results, and present it also for my other lights.
By the way: Thanks a lot to 656nm for the idea.
Earlier this month, I read this tread: Beam Reviews - Update 2/18/07 (added D-mini - also added/changed plots) and got very interested in the suggestion from
656nm at the end of the tread.
Since that, I have been investigating on how to use my camera to get the full output profile from my lights. The idea was to get the lux value in some million points by just one shot.
I painted a wall white with the most diffusive paint I could find, and made some marks as reference points. I put up my L2D at 1 m distance from the wall, took measurements at all reference points, and started to take photos.
I soon found out that it was not that easy as I had hoped. I could only get out the light level in 255 different steps from the photos. This is of course far to little if you want to measure levels of several thousand lux and down to a few. This means i have to take several photos with different exposures.
Another difficulty is that the relationship between light level in the photos and the lux value seems to be linear only for quite low levels. If the light level comes above 100 in the photo (255 is max) then the relationship becomes quite non linear. Regions of the photos with too low light level is difficult to use because a step of one level may mean a big difference in lux level.
I searched around to try to find image processing software that could help me with these problems, and I tried several, but non of them could give me what I wanted, and became very cumbersome to use. So I decided to take up old skills and make the analysis software myself (I worked a lot with programming up to about ten years ago).
I have based my code on the package called ImageMagick and the interface towards the C programming language called ImageWand.
The program I have written, reads in several photos, and compares them with the lux readings from my lux meter. It generates some tables, and writes out a lot of data indicating the accuracy of the conversion of the photos into lux values. There have been a lot of thinking, analysing and trials and error before I could get this to work, but now it does. Here are an example of what I can get out:
This picure is generated out of 9 different images shot with exposure times from 1 s down to 1/1000 seconds. In this picture, every pixel is proportional with the lux value except that I have increased the pixel values at the lower levels just to make them visible. Another alternative is to use a logarithmic scale for the lux levels.
There are a lot of details in this picture that is difficult to see since I have had to reduce to an allowable size. If you see closely, you can se that the rings in the center is skewed slightly to the left. I suspect this is due to specular reflection from my white wall. I will try out other white surfaces than the painted I have now to see if I can get rid of it.
When I am satisfied with my generated lux pictures, then I can take out any profile, and can of cource also very easily calculate the lumens value. I will then figure out the best ways of presenting the results, and present it also for my other lights.
By the way: Thanks a lot to 656nm for the idea.
Last edited: