Abstract: An apparatus converts a multi-density image into a binary value image by converting pixels with densities above a threshold to a first value and pixels below a threshold to a second value. The apparatus generates a histogram relating image densities with the numbers of pixels having those densities. A first memory holds the image pixel densities. A second memory has address inputs attached to the data outputs of the first memory. Addresses of the second memory correspond to image densities. As a pixel is read out of the first memory, its density value serves as an address for a location in the second memory, and a count stored in that address is incremented. After reading all of the pixels from the first memory and incrementing counts, the second memory holds a histogram of the image. The apparatus evaluates the histogram to identify a density for which the number of pixels in the image is a local minimum. The local minimum is selected as the threshold.

Abstract: An image processing method produces a density weighted or unweighted distance transformation of an image. An original image is stored. When a background density is lower than a configuration density, an index image is generated by replacing pixels having a configuration density with a first selected value and other pixels with a zero value. A first image transformed is generated by replacing pixels in the index image with the minimum value of surrounding pixels in a neighborhood. A first addition-fixed image is generated by adding pixels values of the transformed image to corresponding pixel values of the original image. A second transformed image is generated by replacing pixels in the addition-fixed image with a minimum value of surrounding pixels in a neighborhood, and a second addition-fixed image is generated by adding pixel values of the second transformed image to corresponding pixel values of the original image.

Abstract: An image processing method and apparatus generates a concurrence matrix for identifying characteristics in an image. Pixel density values of an original image are stored, and a shifted image is generated by shifting the pixels of the original image by a predetermined distance. A set of counts is generated. Each count corresponds to one density value selected out of the range of possible densities. Each count is the number of times that a pixel in the shifted image equals the selected density simultaneously with a corresponding pixel in the original image equaling a predetermined value. One column in the correspondence matrix corresponds to the counts for one predetermined value. The shift distance is varied to generate a set of matrices. From these matrices, contrast, texture and other characteristics in an image can be quantitatively recognized.

Abstract: A vectorizing method generates an original train of codes for specifying boundary pixels. A revised train of codes is generated solely from the original codes. The revised train may includes points omitted from the original train. Vector points defining boundaries are selected from the revised train of codes.