Text Location In Nature Scene Image

With the development of the people's life, the mobile telephone and digital cameracan get the nature image with word and change the image word into the edited dataword. The localization and extraction of text from nature scene is an important andchallenging research problem in the computer vision. The variation of the text in termsof characters font, size and style, orientation, alignment, texture, color and complexbackground makes the problem of text localization very difficult. The scene content isunconstrained and maybe both indoor and outdoor scenes under anylightingor contrastconditions.To extract nature scene text, first we should find text areas which then subjected toOCR to recognize. Based on the scientific research, we utilize the edge characterizationand texture information to locate text in complicated nature scene. First, it applied edgedetection to R, G, B portion of color image to find maximum edge, then we removedlong lines and noises, used morphologic processing and connected-component analysisto get candidate text regions. Because there are some non-text regions in candidate textregions, then it properly tuned Gabor filter of which frequency parameter selectionutilized Fisher ruling were applied to extract candidate text regions'texture feature, inthe end we sort regions to text regions and non-text regions by varience of filteredimage.The result of the experiment shows that this algorithm is novel and it can well andtrulylocatethetext,andmoreitcan efficientlyreducefalsetextareas. Thisalgorithmisvaluableintheoryandapplication.