摘要 |
Disclosed are a method and a device for extracting a feature for detecting a human based on an image. The method for extracting a feature for detecting a human based on an image comprises the steps of dividing a provided image into blocks of a certain size; calculating a contrast difference value of a center pixel and a peripheral pixel in each of the divided blocks; generating a histogram corresponding to the contrast difference value; generating a feature vector of each of the divided blocks from the histogram; and generating a feature vector of the provided image from the feature vector of the block. Therefore, the feature for detecting a human is extracted to detect the human, thereby reducing detection error among the objects having vertical edges. [Reference numerals] (S110) Divide an image into blocks; (S120) Calculate a contrast difference value of a center pixel and a peripheral pixel in each of the divided blocks; (S130) Accumulate the frequency in histograms in three groups; (S140) Generate a single feature vector for each block; (S150) Generate the single feature vector for all blocks |