LOCALIZATION AND DETECTION OF TEXT REGIONS OF VIDEO SEQUENCES

摘要

The invention disclosed relates to a method and system for localization and detection of text regions of video sequences, the method comprising the following steps: a) compressing the video sequence using an H.264 hybrid video codec; b) calculating the first order differences of DC coefficients of integer transformed residual part of the luma components in the P frames of said compressed video sequence comprising the steps of. i)computing said DC coefficient values for each of the 4x4 sub-blocks obtained from said codec; ii) obtaining said first order differences of said DC coefficient values of neighboring sub-blocks in x and y directions; iii) applying low pass filtering on said obtained first order differences to smoothen the data; iv) defining a threshold value based on which the candidate text regions can be detected; and v) detecting said candidate text regions by comparing the values of said obtained first order differences with said threshold value; c) calculating the vertical edge strengths of each macro block of said compressed video sequence using a deblocking filter; and d) detecting text regions with the help of the values of said first order differences and the values of said vertical edge strengths using a text region detector in a predetermined manner.