The two main papers for the 1/30 class are:

A. Criminisi, I. Reid and A. Zisserman. Single View Metrology, Proc. IEEE International Conference on Computer Vision, 1999

B.M. Oh, M. Chen, J. Dorsey, F. Durand, Image-Based Modeling and Photo Editing. ACM SIGGRAPH 2001.

In the second paper, the authors used filters to differentiate texture and lighting from the original picture (section 5.3). Although it was clear that significant user input was required for the other features in the paper- it wasn't clear to me whether more user input was required in order use the filters.

Would the bi-lateral filter described be sufficient to handle all types of lighting and textures, or does it need to be tweaked for each picture?

Maybe there isn't an answer to this question. Or maybe it will be answered in future papers.

for black-and-white images, intensities between two gray levels are still gray levels. So, smoothing black-white images with simple low-pass Gaussian filtering will cause blurred edges. However, with color images, it will be more complicated. Even in between two colors e.g. blue and red, there will be different
shades of pink and purple. using a bilateral filter---an edge-preserving filter can get much better result. Basic idea is the filter can compute the weight of a pixel by using a Gaussian in the spatial domain to average those pixels with similar intensity values. The method is iterative. My understanding is it doesnt require other specific parameters. you can make one iteration to get a cleaner image, but, multiple iterations have the effect of flattening colors significantly without blurring bounderies in the image.