Test Images

What worked and what didn't

My panoramas taken with the tripod worked pretty well, although there were some problems aligning them perfectly. I believe this was because I was not very careful to ensure that the camera was aligned perfectly vertically. The handheld sequence came out really bad because I was moving around too much while taking it. The last picture didn't line up with the first because of the distance I had moved between the first and last picture.

I didn't use the simple hat-blending scheme. I used an exponential blend function, which seems to work much better. Unfortunately, there's no real notion of blendwidth anymore. You should use a value between 103 and 120 for the parameter that used to be blendwidth. 103 will create a wide blend, 120 will create a narrower blend. It works by dividing the number by 100 and using it as the base of the exponentiation. The exponent is the x distance from the image center-line. The reason it works is that only over a narrow area is the weight of one image similar in magnitude to the weight of the other, since the weight is an exponential function. Thus you don't have to really worry about the blended image areas lining up well. The blend will look correct no matter where the two images happen to line up.

I did four of my panoramas in stitch-assist mode (Downtown, Uptown, Lighthouse, and Parade Grounds) and I did two in manual mode. The ones done in stitch-assist came out with almost no brightness variations between frames, except between the first and last frame because the lighting conditions would usually have changed by the time I got to the last frame. The manual mode ones had horrible brightness variations that I tried to correct in Paint Shop (I hate photoshop). However, the variations are still pretty visible.

Handheld Camera: Sieg Hall, 2nd Floor | Full Image

Test Sequence: University of Washington HUB lawn | Full Image