PURPOSE: To improve the recognition accuracy of an English character recognition device by deciding the propriety of a sentence by collation of words after recognition of characters and applying the post-processing to the sentence after correcting the capital and small letters of same or similar shapes in the case of the sentence.

CONSTITUTION: A sentence deciding part 13 collates the recognition result stored in a recognition result storing part 9 with a word dictionary 11 which previously stores the spellings of English words based on the word areas obtained by a word area segmenting part 5. Thus the part 13 decides whether an image to be recognition object is identical with a sentence or an arrangement of characters. A capital/small letter correcting part 15 corrects the capital and small letters based on a same/similar shape dictionary 14 which stores the same and similar shapes of capital and small letters. Then a post-processing part 10 collates the recognition result stored in the part 9 with the dictionary 11 based on the word areas obtained by the part 5 and corrects the erroneously recognized characters when a sentence is decided by the part 13. Thus the recognition result stored in the part is outputted 9 through a recognition result output part 12.