Codez un logiciel

I want a script or program that I can use in a linux command line to convert images in text. We'll have many images to convert. Images are png or gif format. The images are using only one font, ascii alphabet, and no pictures. On each line, you'll have to recognize and convert to text only the first word. Problems : images are small, font is proportional, and lines are not well seperated (if there is a g in the previous line, it will be seen in the current line zone). you can use an open source ocr like 'tesseract' for example and imagemagick to make images easier to recognize. Warning : using ocr engine on linux command line with standard options, recognition is too bad, you'll have to train or improve image reading via imagemagick for the specific font.