« 19 June 2002 - 3:14 pm | Main | june 20, 2002 - 2:21 piem »

once again its nearing midnight and Im still at work. I think probably people are going to start worrying about me. but where there was bitterness and despair yesterday, today there is hope. first of all, I slept for about 15 hours yesterday after I got home. that helped my outlook a little. also, I am on the verge of a breakthrough in my color balancing problems.

early tests indicate that I should be able to define any color with the function

C = aR + bG + cP

where R, G, and P are reference colors. (uppercase letters denote vectors in uv space and the lowercase letters are scalars.) the only problem is that this equation has multiple solutions, and the only way to derive only one coordinate (a,b,c) is to plot multiple solutions from different locations and find their intersection. still, the fact that there is actually an intersection is very good news. (results were discouraging until I white balanced the images based on a common reference white.)

I need to talk to robert tomorrow and figure out what to do next. my intuition is that I will need to build a look-up table that maps from (u,v) to (a,b,c) as a function of R,G,P for a set of empirically gathered data and then interpolate for all other colors. this probably would work, but it would be time consuming, and so Im hoping someone with more math skills than me will have some insights.

anyway, I spent all day on that, but in the meantime, zach implemented the functions I drew up last week for estimating the (x,y,z) coordinates of a pixel. this is neat, because the robot can now tell how tall I am.

it feels good to finally be making some progress.

(and yeah, I know Im not supposed to post about work, but when you think about something for like 14 hours a day, its kind of hard not.)

Posted by on 19 June 2002 at 11:37 PM