| 主权项 |
1. A method of compensating a signal of an image of one screen for distortions in the horizontal and/or vertical directions of a screen of said image, said method comprising the steps of:
dividing one screen area of said image into a plurality of picture divisions; detecting a movement vector of said image for each particular one of said picture divisions from information on a difference in image between said particular picture division on one of two screens and said particular picture division on the other of said two screens; finding a difference in said detected movement vector between any specific picture division and a picture division adjacent to said specific picture division in order to detect a velocity of a change in image distortion for said specific picture division; and compensating said image for a distortion for each of said picture divisions on the basis of a velocity detected for each of said picture divisions as said velocity of a change in image distortion; wherein said step of detecting a movement vector of said image for each particular one of said picture divisions comprises the sub-steps of: making use of images, which are an observed screen serving as a reference screen and an original screen leading ahead of said reference screen for detecting said movement vector; setting at least one target block having a predetermined size equal to the size of a plurality of pixels at predetermined locations in each of said picture divisions on said original screen; setting a plurality of reference blocks each having a size equal to said target block in a search range set in said reference screen for each of said target blocks; detecting a reference block having the strongest correlation with said specific target block from said plurality of reference blocks; detecting said movement vector representing the magnitude and direction of a shift over said reference screen from a position corresponding to said specific target block to the position of said detected reference block; wherein said movement vector of said image is detected for each of said picture divisions by carrying out: a difference absolute value sum computation process for each specific one of said reference blocks as a process of finding a sum of the absolute values of differences in pixel value between all pixels in said specific reference block and all pixels at corresponding positions on said specific target block; a contracted reference-vector setting process of setting reference vectors each representing the magnitude and direction of a shift over said reference screen from a position corresponding to said specific target block to the position of one of said reference blocks and contracting said reference vectors at a predetermined contraction factor to result in contracted reference vectors; a sum-of-absolute-difference table creation process of creating a shrunk sum-of-absolute-difference table including table elements, the number of which is determined on the basis of the number of said contracted reference vectors and said contraction factor, and allocating each of said table elements as a location to be used for storing a value computed from sums each found for one of said reference blocks as said sum of the absolute values of said differences in pixel values; and a movement-vector computation process of computing said movement vector for each of said picture divisions as a movement vector between said reference screen and said original screen by making use of at least a contracted reference vector corresponding to the smallest one among values each stored in said shrunk sum-of-absolute-difference table as said sum of the absolute values of said differences in pixel values, and said sum-of-absolute-difference table creation process is carried out by execution of a neighborhood reference vector detection sub-process of determining a plurality of neighborhood reference vectors each having a vector quantity close to the vector quantity of a corresponding one of said contracted reference vectors obtained in said contracted reference-vector setting process, a sum-component computation sub-process of computing a plurality of sum components each associated with one of neighborhood reference vectors determined in said neighborhood reference vector detection sub-process as said neighborhood reference vectors corresponding to a specific one of said contracted reference vectors from a sum computed in said difference absolute value sum computation process for a reference block, from which said specific contracted reference vector is contracted, as said sum of the absolute values of said differences in pixel values, and a component-total computation sub-process of computing a total of sum components, which have been computed in said sum-component computation sub-process as said sum components each associated with a specific one of said neighborhood reference vectors, by cumulatively adding each of said sum components to a cumulative total obtained so far for said specific neighborhood reference vector. |
