摘要 |
The invention intends to achieve higher accuracy and higher speeds in image processing based on a stereo method. Images from a plurality cameras are applied to respective image input portions and stored in built-in memories. The stored images are then read out in a certain sequence to cancel lens aberrations of the cameras, and output to respective SAD circuits after being compensated for the differences of intensity values due to variations in the characteristics of the cameras. The SAD circuit calculates an absolute value of the difference for each pixel value between a reference image output from a reference camera and two compared images output from another two cameras, and outputs the calculated value to an SSAD circuit. The SSAD circuit performs a block matching process upon outputs from the corresponding SAD circuit, and outputs the obtained results to a second-order portion. The second-order portion performs second-order interpolation upon the data output from the corresponding SSAD circuit and outputs an obtained result. A minimum value selecting portion and a result memory cooperate to select a minimum value from among the data output from the second-order portions.
|