Training of tesseract

[AKS]

Hi,

I want to use tesseract OCR on images with varying font types and font sizes. Also there is a lot of background variations in each images such as multi-colored background, background with some designs, illumination variance, white background.
If I simply apply  a tesseract with some configuration accroding to my requirement, I am not getting very good accuracy.  Also , due to lot of variation in background and text writing, I cannot apply some fixed image processing on the images before using tesseract.

Can I solve this problem by finetuning tesseract with my set of images??