This thesis presented an algorithm and a prototype program for vehicle license plate recognition from digital images where the license plate consisted of alphabets, numbers, and province name. The font and the format of the license plate are of the standard defined by the Department of Transportation, issue number 25 (1996). The recognition of alphabets, numbers, and province names used a method of dividing alphabets and numbers into three parts horizontally so as lomit the amount of image to be processed. A decision tree was then introduced into the process of identifying the information on the license plate by categorizing space pattern of the data into six types according to the space at the top section and the bottom section as follows, type 1: no space on both upper and lower section, type 2: no space on upper section and one space on lower section, type 3: one space on upper section and no space on lower section, type 4: one space on both upper and lower section, type 5: two spaces on upper section and no space on lower section and type 6: two spaces on upper section and one space on lower section. The outstanding features of all types were then compared. The best match will identify the right alphabet or number.The program gave good results when the image of the license plate was taken with the length of the license plate being about 20% - 50% of the VGA (640 x 480 pixels) image length. The program was tested using 150 images. The program was able to effectively identify the correct location of license plates by 95 percents. The results of the plate location identification is then processed by the license plate recognition step. For the identification of characters and numbers, about 92 percents accuracy was achieved. For the identification of the province names, about 90 percents accuracy was achieved. Thus, the overal accuracy of the license plate recognition was about 89 percents.