Robust detection of video text using an efficient hybrid method via key frame extraction and text localization

Abstract

Video text detection is a challenging problem, since the background of the video image is generally complex and its subtitles often have colour bleeding problems, blurred boundaries and low contrast due to video loss compression and low resolution. Text detection is an important method for many image processing tasks that are focused on text. In this paper, we put forward a robust detection method for extracting video text using hybrid method of MSER via morphological filtering for solving these problems. This can also solve the problems of bleeding in colour and floured boundaries. In this we added 2-D DWT (discrete wavelet transforms) is developed to remove background noise and improve sound and text contrast. SO that components are extracted with MSER from origin and processed images. In this work, the proposed method develops an efficient method of extracting and recognizing text, using the principle of morphological operations using MATLAB. Current text extraction methods–edge dependent and connected components when implemented separately yield better results. But using these approaches sometimes cannot get better results as well as its time taken. Therefore it is suggested that combine both methods, the outcome shows that the approach suggested produces better results than the other two approaches.