Abstarct: Word spotting is a way to search for text information in document images. This process includes finding and locating a query word within a document image. In this context, more researches are provided on Latin documents. A Few researches in Arabic and Farsi has been reported, the majority of these researches are on printed documents. In this thesis we present a Farsi handwritten word spotting system. The proposed system is baxsed on two steps. In the first stage, all existing connected components in the text image have been extracted, then the first connected component of each line is determined. Therefore by applying the nearest neighbor method, all the connected components of each line obtained according to their order in the text. As a result, after this stage, each of the connected components in the text devoted to one of the lines. Actually each line can be demonstrated with its connected components independent of other lines. The result of this stage is decomposing the text image into separated lines. In the next step with regard to the rule that the majority of prepositions and Persian words are combination of one to seven connected components, the words in the text are extracted. Actually the connected components of an independent word in the text are indentified in the same order of appearance in the word. For each of the word images extracted, the coordinates of their location in the text is saved. As a result, after the first stage a set of candidate words is obtained. In the second stage, the goal is to find all instances of a query word in this set then their location in the text is specified. The query word may be a text string or may also be an image. In this stage, for word spotting we use attributes from Persian alphabet. The use of this kind of attributes makes word spotting independent of the writing style. This approach has been applied on a dataset consisting of 55 query words and 100 Farsi handwritten document images. Average precision and recall for query by image is 75.45% and 80.42% and for query by text is 80.28% and 78.92%.