| 摘要 |
PURPOSE:To easily and rapidly execute processing by smoothing the density histogram of image data obtained by quantizing the color density of red, green and blue and identifying the attribution of an image on the basis of the number of feature points extracted from the smoothed density histogram CONSTITUTION:The number (frequency) of picture elements having the same color density is accumulated from inputted image data to form a density histogram, the frequencies of the preceding and succeeding several levels of each density level are added, the added value is divided by the value of the added density level to smooth the frequency in each density level. The obtained histogram curve is stored as linear graphic data and broken line approximating processing is applied to the histogram curve on the basis of the linear graphic data. Bent point obtained as the result of the broken line approximating processing are extracted as feature points and a multi-color image and a full color image are identified n the basis of the number of feature points. Consequently, the number of data bits per picture element can be reduced, the processing based upon a microcomputer can be easily executed and image processing can be rapidly executed. |
